Joseph O. Deasy

Deformable image registration is widely used in various radiation therapy applications including daily treatment planning adaptation to map planned tissue or dose to changing anatomy. In this work, a simple and efficient inverse consistency deformable registration method is proposed with aims of higher registration accuracy and faster convergence speed. Instead of registering image I to a second image J, the two images are symmetrically deformed toward one another in multiple passes, until both deformed images are matched and correct registration is therefore achieved. In each pass, a delta motion field is computed by minimizing a symmetric optical flow system cost function using modified optical flow algorithms. The images are then further deformed with the delta motion field in the positive and negative directions respectively, and then used for the next pass. The magnitude of the delta motion field is forced to be less than 0.4 voxel for every pass in order to guarantee smoothness and invertibility for the two overall motion fields that are accumulating the delta motion fields in both positive and negative directions, respectively. The final motion fields to register the original images I and J, in either direction, are calculated by inverting one overall motion field and combining the inversion result with the other overall motion field. The final motion fields are inversely consistent and this is ensured by the symmetric way that registration is carried out. The proposed method is demonstrated with phantom images, artificially deformed patient images and 4D-CT images. Our results suggest that the proposed method is able to improve the overall accuracy (reducing registration error by 30% or more, compared to the original and inversely inconsistent optical flow algorithms), reduce the inverse consistency error (by 95% or more) and increase the convergence rate (by 100% or more). The overall computation speed may slightly decrease, or increase in most cases because the new method converges faster. Compared to previously reported inverse consistency algorithms, the proposed method is simpler, easier to implement and more efficient.

Abstract Building surfaces from cross‐section curves has wide applications including bio‐medical modeling. Previous work in this area has mostly focused on connecting simple closed curves on parallel cross‐sections. Here we consider the more general problem where input data may lie on non‐parallel cross‐sections and consist of curve networks that represent the segmentation of the underlying object by different material or tissue types (e.g., skin, muscle, bone, etc.) on each cross‐section. The desired output is a surface network that models both the exterior surface and the internal partitioning of the object. We introduce an algorithm that is capable of handling curve networks of arbitrary shape and topology on cross‐section planes with arbitrary orientations. Our algorithm is simple to implement and is guaranteed to produce a closed surface network that interpolates the curve network on each cross‐section. Our method is demonstrated on both synthetic and bio‐medical examples.

Introduction Acute dysphagia is a distressing dose-limiting toxicity occurring frequently during concurrent chemo-radiation or high-dose radiotherapy for lung cancer. It can lead to treatment interruptions and thus jeopardize survival. Although a number of predictive factors have been identified, it is still not clear how these could offer assistance for treatment decision making in daily clinical practice. Therefore, we have developed and validated a nomogram to predict this side-effect. In addition, clinical usefulness was assessed by comparing model predictions to physicians’ predictions. Materials and methods Clinical data from 469 inoperable lung cancer patients, treated with curative intent, were collected prospectively. A prediction model for acute radiation-induced dysphagia was developed. Model performance was evaluated by the c-statistic and assessed using bootstrapping as well as two external datasets. In addition, a prospective study was conducted comparing model to physicians’ predictions in 138 patients. Results The final multivariate model consisted of age, gender, WHO performance status, mean esophageal dose (MED), maximum esophageal dose (MAXED) and overall treatment time (OTT). The c-statistic, assessed by bootstrapping, was 0.77. External validation yielded an AUC of 0.94 on the Ghent data and 0.77 on the Washington University St. Louis data for dysphagia ⩾ grade 3. Comparing model predictions to the physicians’ predictions resulted in an AUC of 0.75 versus 0.53, respectively. Conclusions The proposed model performed well was successfully validated and demonstrated the ability to predict acute severe dysphagia remarkably better than the physicians. Therefore, this model could be used in clinical practice to identify patients at high or low risk.

Emerging data suggest that dose-sparing several key cardiac regions is prognostically beneficial in lung cancer radiotherapy. The cardiac substructures are challenging to contour due to their complex geometry, poor soft tissue definition on computed tomography (CT) and cardiorespiratory motion artefact. A neural network was previously trained to generate the cardiac substructures using three-dimensional radiotherapy planning CT scans (3D-CT). In this study, the performance of that tool on the average intensity projection from four-dimensional (4D) CT scans (4D-AVE), now commonly used in lung radiotherapy, was evaluated.The 4D-AVE of n=20 patients completing radiotherapy for lung cancer 2015-2020 underwent manual and automated cardiac substructure segmentation. Manual and automated substructures were compared geometrically and dosimetrically. Two senior clinicians also qualitatively assessed the auto-segmentation tool's output.Geometric comparison of the automated and manual segmentations exhibited high levels of similarity across parameters, including volume difference (11.8% overall) and Dice similarity coefficient (0.85 overall), and were consistent with 3D-CT performance. Differences in mean (median 0.2 Gy, range -1.6-0.3 Gy) and maximum (median 0.4 Gy, range -2.2-0.9 Gy) doses to substructures were generally small. Nearly all structures (99.5 %) were deemed to be appropriate for clinical use without further editing.Cardiac substructure auto-segmentation using a deep learning-based tool trained on a 3D-CT dataset was feasible on the 4D-AVE scan, meaning this tool is suitable for use on 4D-CT radiotherapy planning scans. Application of this tool would increase the practicality of routine clinical cardiac substructure delineation, and enable further cardiac radiation effects research.

To investigate the merits of texture analysis on parametric maps derived from pharmacokinetic modeling with dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) as imaging biomarkers for the prediction of treatment response in patients with head and neck squamous cell carcinoma (HNSCC).In this retrospective study, 19 HNSCC patients underwent pre- and intra-treatment DCE-MRI scans at a 1.5T MRI scanner. All patients had chemo-radiation treatment. Pharmacokinetic modeling was performed on the acquired DCE-MRI images, generating maps of volume transfer rate (K(trans)) and volume fraction of the extravascular extracellular space (ve). Image texture analysis was then employed on maps of K(trans) and ve, generating two texture measures: Energy (E) and homogeneity.No significant changes were found for the mean and standard deviation for K(trans) and ve between pre- and intra-treatment (P > 0.09). Texture analysis revealed that the imaging biomarker E of ve was significantly higher in intra-treatment scans, relative to pretreatment scans (P < 0.04).Chemo-radiation treatment in HNSCC significantly reduces the heterogeneity of tumors.

Purpose: Develop an integrated intra-site and inter-site radiomics-clinical-genomic marker of high grade serous ovarian cancer (HGSOC) outcomes and explore the biological basis of radiomics with respect to molecular signaling pathways and the tumor microenvironment (TME). Method: Seventy-five stage III-IV HGSOC patients from internal (N = 40) and external factors via the Cancer Imaging Archive (TCGA) (N = 35) with pre-operative contrast enhanced CT, attempted primary cytoreduction, at least two disease sites, and molecular analysis performed within TCGA were retrospectively analyzed. An intra-site and inter-site radiomics (cluDiss) measure was combined with clinical-genomic variables (iRCG) and compared against conventional (volume and number of sites) and average radiomics (N = 75) for prognosticating progression-free survival (PFS) and platinum resistance. Correlation with molecular signaling and TME derived using a single sample gene set enrichment that was measured. Results: The iRCG model had the best platinum resistance classification accuracy (AUROC of 0.78 [95% CI 0.77 to 0.80]). CluDiss was associated with PFS (HR 1.03 [95% CI: 1.01 to 1.05], p = 0.002), negatively correlated with Wnt signaling, and positively to immune TME. Conclusions: CluDiss and the iRCG prognosticated HGSOC outcomes better than conventional and average radiomic measures and could better stratify patient outcomes if validated on larger multi-center trials.

The development of reliable predictive models for individual cancer cell lines to identify an optimal cancer drug is a crucial step to accelerate personalized medicine, but vast differences in cancer cell lines and drug characteristics make it quite challenging to develop predictive models that result in high predictive power and explain the similarity of cell lines or drugs. Our study proposes a novel network-based methodology that breaks the problem into smaller, more interpretable problems to improve the predictive power of anti-cancer drug responses in cell lines. For the drug-sensitivity study, we used the GDSC database for 915 cell lines and 200 drugs. The theory of optimal mass transport was first used to separately cluster cell lines and drugs, using gene-expression profiles and extensive cheminformatic drug features, represented in a form of data networks. To predict cell-line specific drug responses, random forest regression modeling was separately performed for each cell-line drug cluster pair. Post-modeling biological analysis was further performed to identify potential biological correlates associated with drug responses. The network-based clustering method resulted in 30 distinct cell-line drug cluster pairs. Predictive modeling on each cell-line-drug cluster outperformed alternative computational methods in predicting drug responses. We found that among the four drugs top-ranked with respect to prediction performance, three targeted the PI3K/mTOR signaling pathway. Predictive modeling on clustered subsets of cell lines and drugs improved the prediction accuracy of cell-line specific drug responses. Post-modeling analysis identified plausible biological processes associated with drug responses.

Abstract Background Fast and accurate multiorgans segmentation from computed tomography (CT) scans is essential for radiation treatment planning. Self‐attention(SA)‐based deep learning methodologies provide higher accuracies than standard methods but require memory and computationally intensive calculations, which restricts their use to relatively shallow networks. Purpose Our goal was to develop and test a new computationally fast and memory‐efficient bidirectional SA method called nested block self‐attention (NBSA), which is applicable to shallow and deep multiorgan segmentation networks. Methods A new multiorgan segmentation method combining a deep multiple resolution residual network with computationally efficient SA called nested block SA (MRRN‐NBSA) was developed and evaluated to segment 18 different organs from head and neck (HN) and abdomen organs. MRRN‐NBSA combines features from multiple image resolutions and feature levels with SA to extract organ‐specific contextual features. Computational efficiency is achieved by using memory blocks of fixed spatial extent for SA calculation combined with bidirectional attention flow. Separate models were trained for HN ( n = 238) and abdomen ( n = 30) and tested on set aside open‐source grand challenge data sets for HN ( n = 10) using a public domain database of computational anatomy and blinded testing on 20 cases from Beyond the Cranial Vault data set with overall accuracy provided by the grand challenge website for abdominal organs. Robustness to two‐rater segmentations was also evaluated for HN cases using the open‐source data set. Statistical comparison of MRRN‐NBSA against Unet, convolutional network–based SA using criss‐cross attention (CCA), dual SA, and transformer‐based (UNETR) methods was done by measuring the differences in the average Dice similarity coefficient (DSC) accuracy for all HN organs using the Kruskall–Wallis test, followed by individual method comparisons using paired, two‐sided Wilcoxon‐signed rank tests at 95% confidence level with Bonferroni correction used for multiple comparisons. Results MRRN‐NBSA produced an average high DSC of 0.88 for HN and 0.86 for the abdomen that exceeded current methods. MRRN‐NBSA was more accurate than the computationally most efficient CCA (average DSC of 0.845 for HN, 0.727 for abdomen). Kruskal–Wallis test showed significant difference between evaluated methods ( p =0.00025). Pair‐wise comparisons showed significant differences between MRRN‐NBSA than Unet ( p =0.0003), CCA ( p =0.030), dual ( p =0.038), and UNETR methods ( p =0.012) after Bonferroni correction. MRRN‐NBSA produced less variable segmentations for submandibular glands (0.82 ± 0.06) compared to two raters (0.75 ± 0.31). Conclusions MRRN‐NBSA produced more accurate multiorgan segmentations than current methods on two different public data sets. Testing on larger institutional cohorts is required to establish feasibility for clinical use.

This study was conducted to identify the factors associated with acute gastrointestinal (GI) toxicity during pelvic chemoradiotherapy (PCRT) in patients with rectal cancer.We analyzed 177 patients with rectal cancer treated from 2007 through 2010. Clinical information, including weekly diarrhea and proctitis toxicity grade during PCRT, was recorded. GI structures including bowel and anal canal were contoured. The associations between toxicity and clinical and dosimetric predictors were tested.The median age was 60; 76 patients were women; 98 were treated with intensity-modulated radiotherapy (IMRT) and 79 with 3D conformal RT (3DCRT). A higher rate of grade 2+ diarrhea was observed in the women, starting at week 4 (24% women vs. 11% men, P = .01; week 5: 33% vs. 12%, P = .002), as well as in all the patients treated with 3DCRT (22% vs. 12% IMRT, P = .03; week 5: 32% vs. 11%, P = .001). On multivariate analysis, the normal tissue complication probability (NTCP) model including bowel V45 (bowel volume receiving ≥45 Gy) showed that being female, and use of 3DCRT, was most predictive of grade 2+ diarrhea (area under the curve [AUC] = 0.76; R S = 0.35; P < .001). A higher rate of grade 2+ proctitis was seen in patients <60 years of age starting at week 3 (21% vs. 9%, P = .02; week 4: 35% vs. 16%, P = .003). The NTCP model including anal canal V15 and younger age was most predictive of grade 2+ proctitis (AUC = 0.67; R S = 0.25; P < .001).Women and all patients who were treated with 3DCRT had higher rates of grade 2+ diarrhea, and the younger patients had a higher rate of grade 2+ proctitis during PCRT. The use of more stringent dosimetric constraints in higher risk patients is a strategy for minimizing toxicity.

Data visualization technique was applied to analyze the daily QA results of photon and electron beams.Special attention was paid to any trend the beams might display.A Varian Trilogy Linac equipped with dual photon energies and five electron energies was commissioned in early 2010.Daily Linac QA tests including the output constancy, beam flatness and symmetry (radial and transverse directions) were performed with an ionization chamber array device (QA Beam Checker Plus, Standard Imaging).The data of five years were collected and analyzed.For each energy, the measured data were exported and processed for visual trending using an in-house Matlab program.These daily data were cross-correlated with the monthly QA and annual QA results, as well as the preventive maintenance records.Majority of the output were within 1% of variation, with a consistent positive/upward drift for all seven energies (~+0.25% per month).The baseline of daily device is reset annually right after the TG-51 calibration.This results in a sudden drop of the output.On the other hand, the large amount of data using the same baseline exhibits a sinusoidal behavior (period = 12 months; amplitude = 0.8%, 0.5% for photons, electrons, respectively) on symmetry and flatness when normalization of baselines is accounted for.The well known phenomenon of new Linac output drift was clearly displayed.This output drift was a result of the air leakage of the over-pressurized sealed monitor chambers for the specific vendor.Data visualization is a new trend in the era of big data in radiation oncology research.It allows the data to be displayed visually and therefore more intuitive.Based on the visual display from the past, the physicist might predict the trend of the Linac and take actions proactively.It also makes comparisons, alerts failures, and potentially identifies causalities.

Each year, approximately 18 million new cancer cases are diagnosed worldwide, and about half must be treated with radiotherapy. A successful treatment requires treatment planning with the customization of penetrating radiation beams to sterilize cancerous cells without harming nearby normal organs and tissues. This process currently involves extensive manual tuning of parameters by an expert planner, making it a time-consuming and labor-intensive process, with quality and immediacy of critical care dependent on the planner’s expertise. To improve the speed, quality, and availability of this highly specialized care, Memorial Sloan Kettering Cancer Center developed and applied advanced optimization tools to this problem (e.g., using hierarchical constrained optimization, convex approximations, and Lagrangian methods). This resulted in both a greatly improved radiotherapy treatment planning process and the generation of reliable and consistent high-quality plans that reflect clinical priorities. These improved techniques have been the foundation of high-quality treatments and have positively impacted over 5,000 patients to date, including numerous patients in severe pain and in urgent need of treatment who might have otherwise required longer hospital stays or undergone unnecessary surgery to control the progression of their disease. We expect that the wide distribution of the system we developed will ultimately impact patient care more broadly, including in resource-constrained countries.

Abstract To identify molecular targets that would be missed by more conventional methods of analysis, we applied a novel unsupervised method to cluster gene interactomes represented as networks in pediatric sarcoma. RNA-Seq data were mapped to protein-level interactomes to construct weighted networks that were then subjected to a non-Euclidean, multi-scale geometric approach centered on a discrete notion of curvature. This provides a measure of the functional association between genes in the context of their connectivity. In confirmation of the validity of this method, hierarchical clustering revealed the characteristic EWSR1-FLI1 fusion in Ewing sarcoma. Furthermore, assessing the effects of in silico edge perturbations and simulated gene knockouts as quantified by changes in curvature, we found non-trivial gene associations not previously identified.

Gastrointestinal (GI) tract motility is one of the main sources for intra/inter-fraction variability and uncertainty in radiation therapy for abdominal targets. Models for GI motility can improve the assessment of delivered dose and contribute to the development, testing, and validation of deformable image registration (DIR) and dose-accumulation algorithms.To implement GI tract motion in the 4D extended cardiac-torso (XCAT) digital phantom of human anatomy.Motility modes that exhibit large amplitude changes in the diameter of the GI tract and may persist over timescales comparable to online adaptive planning and radiotherapy delivery were identified based on literature research. Search criteria included amplitude changes larger than planning risk volume expansions and durations of the order of tens of minutes. The following modes were identified: peristalsis, rhythmic segmentation, high amplitude propagating contractions (HAPCs), and tonic contractions. Peristalsis and rhythmic segmentations were modeled by traveling and standing sinusoidal waves. HAPCs and tonic contractions were modeled by traveling and stationary Gaussian waves. Wave dispersion in the temporal and spatial domain was implemented by linear, exponential, and inverse power law functions. Modeling functions were applied to the control points of the nonuniform rational B-spline surfaces defined in the reference XCAT library. GI motility was combined with the cardiac and respiratory motions available in the standard 4D-XCAT phantom. Default model parameters were estimated based on the analysis of cine MRI acquisitions in 10 patients treated in a 1.5T MR-linac.We demonstrate the ability to generate realistic 4D multimodal images that simulate GI motility combined with respiratory and cardiac motion. All modes of motility, except tonic contractions, were observed in the analysis of our cine MRI acquisitions. Peristalsis was the most common. Default parameters estimated from cine MRI were used as initial values for simulation experiments. It is shown that in patients undergoing stereotactic body radiotherapy for abdominal targets, the effects of GI motility can be comparable or larger than the effects of respiratory motion.The digital phantom provides realistic models to aid in medical imaging and radiation therapy research. The addition of GI motility will further contribute to the development, testing, and validation of DIR and dose accumulation algorithms for MR-guided radiotherapy.

Cancer transcriptional patterns exhibit both shared and unique features across diverse cancer types, but whether these patterns are sufficient to characterize the full breadth of tumor phenotype heterogeneity remains an open question. We hypothesized that cancer transcriptional diversity mirrors patterns in normal tissues optimized for distinct functional tasks. Starting with normal tissue transcriptomic profiles, we use non-negative matrix factorization to derive six distinct transcriptomic phenotypes, called archetypes, which combine to describe both normal tissue patterns and variations across a broad spectrum of malignancies. We show that differential enrichment of these signatures correlates with key tumor characteristics, including overall patient survival and drug sensitivity, independent of clinically actionable DNA alterations. Additionally, we show that in HR+/HER2-breast cancers, metastatic tumors adopt transcriptomic signatures consistent with the invaded tissue. Broadly, our findings suggest that cancer often arrogates normal tissue transcriptomic characteristics as a component of both malignant progression and drug response. This quantitative framework provides a strategy for connecting the diversity of cancer phenotypes and could potentially help manage individual patients.

To determine the mechanisms involved in partial volume radiation therapy (RT)-induced tumor response.We investigated 67NR murine orthotopic breast tumors in Balb/c mice and Lewis lung carcinoma (LLC cells; WT, Crispr/Cas9 Sting KO, and Atm KO) injected in the flank of C57Bl/6, cGAS, or STING KO mice. RT was delivered to 50% or 100% of the tumor volume using a 2 × 2 cm collimator on a microirradiator allowing precise irradiation. Tumors and blood were collected at 6, 24, and 48 hours post-RT and assessed for cytokine measurements.There is a significant activation of the cGAS/STING pathway in the hemi-irradiated tumors compared with control and to 100% exposed 67NR tumors. In the LLC model, we determined that an ATM-mediated noncanonical activation of STING is involved. We demonstrated that the partial exposure RT-mediated immune response is dependent on ATM activation in the tumor cells and on the STING activation in the host, and cGAS is dispensable. Our results also indicate that partial volume RT stimulates a proinflammatory cytokine response compared with the anti-inflammatory profile induced by 100% tumor volume exposure.Partial volume RT induces an antitumor response by activating STING, which stimulates a specific cytokine signature as part of the immune response. However, the mechanism of this STING activation, via the canonical cGAS/STING pathway or a noncanonical ATM-driven pathway, depends on the tumor type. Identifying the upstream pathways responsible for STING activation in the partial RT-mediated immune response in different tumor types would improve this therapy and its potential combination with immune checkpoint blockade and other antitumor therapies.

Patients with brain metastases (BMs) are surviving longer and returning for multiple courses of stereotactic radiosurgery. BMs are monitored after radiation with follow-up magnetic resonance (MR) imaging every 2-3 months. This study investigated whether it is possible to automatically track BMs on longitudinal imaging and quantify the tumor response after radiotherapy.The METRO process (MEtastasis Tracking with Repeated Observations was developed to automatically process patient data and track BMs. A longitudinal intrapatient registration method for T1 MR post-Gd was conceived and validated on 20 patients. Detections and volumetric measurements of BMs were obtained from a deep learning model. BM tracking was validated on 32 separate patients by comparing results with manual measurements of BM response and radiologists' assessments of new BMs. Linear regression and residual analysis were used to assess accuracy in determining tumor response and size change.A total of 123 irradiated BMs and 38 new BMs were successfully tracked. 66 irradiated BMs were visible on follow-up imaging 3-9 months after radiotherapy. Comparing their longest diameter changes measured manually vs. METRO, the Pearson correlation coefficient was 0.88 (p < 0.001); the mean residual error was -8 ± 17%. The mean registration error was 1.5 ± 0.2 mm.Automatic, longitudinal tracking of BMs using deep learning methods is feasible. In particular, the software system METRO fulfills a need to automatically track and quantify volumetric changes of BMs prior to, and in response to, radiation therapy.

Disease progression after definitive stereotactic body radiation therapy (SBRT) for early-stage non-small cell lung cancer (NSCLC) occurs in 20–40% of patients. Here, we explored published and novel pre-treatment CT and PET radiomics features to identify patients at risk of progression. Published CT and PET features were identified and explored along with 15 other CT and PET features in 408 consecutively treated early-stage NSCLC patients having CT and PET < 3 months pre-SBRT (training/set-aside validation subsets: n = 286/122). Features were associated with progression-free survival (PFS) using bootstrapped Cox regression (Bonferroni-corrected univariate predictor: p ≤ 0.002) and only non-strongly correlated predictors were retained (|Rs|<0.70) in forward-stepwise multivariate analysis. Tumor diameter and SUVmax were the two most frequently reported features associated with progression/survival (in 6/20 and 10/20 identified studies). These two features and 12 of the 15 additional features (CT: 6; PET: 6) were candidate PFS predictors. A re-fitted model including diameter and SUVmax presented with the best performance (c-index: 0.78; log-rank p-value < 0.0001). A model built with the two best additional features (CTspiculation1 and SUVentropy) had a c-index of 0.75 (log-rank p-value < 0.0001). A re-fitted pre-treatment model using the two most frequently published features – tumor diameter and SUVmax – successfully stratified early-stage NSCLC patients by PFS after receiving SBRT.

Abstract Network properties account for the complex relationship between genes, making it easier to identify complex patterns in their interactions. In this work, we leveraged these network properties for dual purposes. First, we clustered pediatric sarcoma tumors using network information flow as a similarity metric, computed by the Wasserstein distance. We demonstrate that this approach yields the best concordance with histological subtypes, validated against three state-of-the-art methods. Second, to identify molecular targets that would be missed by more conventional methods of analysis, we applied a novel unsupervised method to cluster gene interactomes represented as networks in pediatric sarcoma. RNA-Seq data were mapped to protein-level interactomes to construct weighted networks that were then subjected to a non-Euclidean, multi-scale geometric approach centered on a discrete notion of curvature. This provides a measure of the functional association among genes in the context of their connectivity. In confirmation of the validity of this method, hierarchical clustering revealed the characteristic EWSR1 - FLI1 fusion in Ewing sarcoma. Furthermore, assessing the effects of in silico edge perturbations and simulated gene knockouts as quantified by changes in curvature, we found non-trivial gene associations not previously identified.

Objective assessment of delivered radiotherapy (RT) to thoracic organs requires fast and accurate deformable dose mapping. The aim of this study was to implement and evaluate an artificial intelligence (AI) deformable image registration (DIR) and organ segmentation-based AI dose mapping (AIDA) applied to the esophagus and the heart.AIDA metrics were calculated for 72 locally advanced non-small cell lung cancer patients treated with concurrent chemo-RT to 60 Gy in 2 Gy fractions in an automated pipeline. The pipeline steps were: (i) automated rigid alignment and cropping of planning CT to week 1 and week 2 cone-beam CT (CBCT) field-of-views, (ii) AI segmentation on CBCTs, and (iii) AI-DIR-based dose mapping to compute dose metrics. AIDA dose metrics were compared to the planned dose and manual contour dose mapping (manual DA).AIDA required ∼2 min/patient. Esophagus and heart segmentations were generated with a mean Dice similarity coefficient (DSC) of 0.80±0.15 and 0.94±0.05, a Hausdorff distance at 95th percentile (HD95) of 3.9±3.4 mm and 14.1±8.3 mm, respectively. AIDA heart dose was significantly lower than the planned heart dose (p = 0.04). Larger dose deviations (>=1Gy) were more frequently observed between AIDA and the planned dose (N = 26) than with manual DA (N = 6).Rapid estimation of RT dose to thoracic tissues from CBCT is feasible with AIDA. AIDA-derived metrics and segmentations were similar to manual DA, thus motivating the use of AIDA for RT applications.

Radiation therapy dose distributions to eradicate tumor cells are typically constrained in extent or intensity to minimize the risk of injury to nearby critical normal tissues. With the widespread use of 3D image-based treatment planning systems, the question naturally arises how patient-specific anatomy and treatment differences affect outcome. It has long been known, that for many organs, variations in the fractional volume irradiated to high doses greatly alters the dose to achieve a given complication level (the “isoeffective dose”) [1]. Smaller irradiated fractional volumes often lead to a much lower risk of complication; this is often referred to as the “volume-effect” in the literature, but would be more correctly referred to as the “dose–volume” effect. Normal tissue complication probability (NTCP) modeling is simply the ongoing effort to understand the risk of normal tissue injury as a function of the 3D dose distribution.

To automatically estimate average diaphragm motion trajectory (ADMT) based on four-dimensional computed tomography (4DCT), facilitating clinical assessment of respiratory motion and motion variation and retrospective motion study. We have developed an effective motion extraction approach and a machine-learning-based algorithm to estimate the ADMT. Eleven patients with 22 sets of 4DCT images (4DCT1 at simulation and 4DCT2 at treatment) were studied. After automatically segmenting the lungs, the differential volume-per-slice (dVPS) curves of the left and right lungs were calculated as a function of slice number for each phase with respective to the full-exhalation. After 5-slice moving average was performed, the discrete cosine transform (DCT) was applied to analyze the dVPS curves in frequency domain. The dimensionality of the spectrum data was reduced by using several lowest frequency coefficients (fv) to account for most of the spectrum energy (Σfv2). Multiple linear regression (MLR) method was then applied to determine the weights of these frequencies by fitting the ground truth-the measured ADMT, which are represented by three pivot points of the diaphragm on each side. The 'leave-one-out' cross validation method was employed to analyze the statistical performance of the prediction results in three image sets: 4DCT1, 4DCT2, and 4DCT1 + 4DCT2. Seven lowest frequencies in DCT domain were found to be sufficient to approximate the patient dVPS curves (R = 91%-96% in MLR fitting). The mean error in the predicted ADMT using leave-one-out method was 0.3 ± 1.9 mm for the left-side diaphragm and 0.0 ± 1.4 mm for the right-side diaphragm. The prediction error is lower in 4DCT2 than 4DCT1, and is the lowest in 4DCT1 and 4DCT2 combined. This frequency-analysis-based machine learning technique was employed to predict the ADMT automatically with an acceptable error (0.2 ± 1.6 mm). This volumetric approach is not affected by the presence of the lung tumors, providing an automatic robust tool to evaluate diaphragm motion.

The era of noninvasive diagnostic radiology and image-guided radiotherapy has witnessed burgeoning interest in applying different imaging modalities to stage and localize complex diseases such as atherosclerosis or cancer. It has been observed that using complementary information from multimodality images often significantly improves the robustness and accuracy of target volume definitions in radiotherapy treatment of cancer. In this work, we present techniques and an interactive software tool to support this new framework for 3D multimodality medical image segmentation. To demonstrate this methodology, we have designed and developed a dedicated open source software tool for multimodality image analysis MIASYS. The software tool aims to provide a needed solution for 3D image segmentation by integrating automatic algorithms, manual contouring methods, image preprocessing filters, post-processing procedures, user interactive features and evaluation metrics. The presented methods and the accompanying software tool have been successfully evaluated for different radiation therapy and diagnostic radiology applications.

To noninvasively investigate tumor cellularity measured using diffusion-weighted magnetic resonance imaging (DW-MRI) and glucose metabolism measured by 18F-labeled fluorodeoxyglucose positron emission tomography/computed tomography (18F-FDG-PET/CT) during radiation therapy (RT) for human papillomavirus negative (HPV-) head and neck squamous cell carcinoma (HNSCC).In this prospective study, 6 HPV- HNSCC patients underwent a total of 34 multimodality imaging examinations DW-MRI at 1.5 T Philips MRI scanner [(n = 24) pre-, during- (2-3 wk), and post-treatment (Tx), and 18F-FDG PET/CT pre- and post-Tx (n = 10)]. All patients received RT. Monoexponential modeling of the DW-MRI data yielded the imaging metric apparent diffusion coefficient (ADC) and the mean of standardized uptake value (SUV) was measured from 18F-FDG PET uptake. All patients had a clinical follow-up as the standard of care and survival status was documented at 1 year.There was a strong negative correlation between the mean of pretreatment ADC (ρ = -0.67, P = 0.01) and the pretreatment 18F-FDG PET SUV. The percentage (%) change in delta (∆) ADC for primary tumors and neck nodal metastases between pre- and Wk2-3 Tx were as follows: 75.4% and 61.6%, respectively, for the patient with no evidence of disease, 27.5% and 32.7%, respectively, for those patients who were alive with disease, and 26.9% and 7.31%, respectively, for those who were dead with disease.These results are preliminary in nature and are indicative, and not definitive, trends rendered by the imaging metrics due to the small sample size of HPV- HNSCC patients in a Meixoeiro Hospital of Vigo Experience.

In this paper, we present a network-based clustering method, called vector Wasserstein clustering (vWCluster), based on the vector-valued Wasserstein distance derived from optimal mass transport (OMT) theory. This approach allows for the natural integration of multi-layer representations of data in a given network from which one derives clusters via a hierarchical clustering approach. In this study, we applied the methodology to multi-omics data from the two largest breast cancer studies. The resultant clusters showed significantly different survival rates in Kaplan-Meier analysis in both datasets. CIBERSORT scores were compared among the identified clusters. Out of the 22 CIBERSORT immune cell types, 9 were commonly significantly different in both datasets, suggesting the difference of tumor immune microenvironment in the clusters. vWCluster can aggregate multi-omics data represented as a vectorial form in a network with multiple layers, taking into account the concordant effect of heterogeneous data, and further identify subgroups of tumors in terms of mortality.

Compositional systems, represented as parts of some whole, are ubiquitous. They encompass the abundances of proteins in a cell, the distribution of organisms in nature, and the stoichiometry of the most basic chemical reactions. Thus, a central goal is to understand how such processes emerge from the behaviors of their components and their pairwise interactions. Such a study, however, is challenging for two key reasons. Firstly, such systems are complex and depend, often stochastically, on their constituent parts. Secondly, the data lie on a simplex which influences their correlations.To resolve both of these issues, we provide a general and data-driven modeling tool for compositional systems called Compositional Maximum Entropy (CME). By integrating the prior geometric structure of compositions with sample-specific information, CME infers the underlying multivariate relationships between the constituent components. We provide two proofs of principle. First, we measure the relative abundances of different bacteria and infer how they interact. Second, we show that our method outperforms a common alternative for the extraction of gene-gene interactions in triple-negative breast cancer.CME provides novel and biologically-intuitive insights and is promising as a comprehensive quantitative framework for compositional data.

ABSTRACT The plasma cell cancer multiple myeloma (MM) varies significantly in genomic characteristics, response to therapy, and long-term prognosis. To investigate global interactions in MM, we combined a known protein interaction network with a large clinically annotated MM dataset. We hypothesized that an unbiased network analysis method based on large-scale similarities in gene expression, copy number aberration, and protein interactions may provide novel biological insights. Applying a novel measure of network robustness, Ollivier-Ricci Curvature, we examined patterns in the RNA-Seq gene expression and CNA data and how they relate to clinical outcomes. Hierarchical clustering using ORC differentiated high-risk subtypes with low progression free survival. Differential gene expression analysis defined 118 genes with significantly aberrant expression. These genes, while not previously associated with MM, were associated with DNA repair, apoptosis, and the immune system. Univariate analysis identified 8/118 to be prognostic genes; all associated with the immune system. A network topology analysis identified both hub and bridge genes which connect known genes of biological significance of MM. Taken together, gene interaction network analysis in MM uses a novel method of global assessment to demonstrate complex immune dysregulation associated with shorter survival. STATEMENT OF SIGNIFICANCE Multiple myeloma has heterogenous clinical outcomes which are not well predicted by current prognostic scoring systems. Global assessment of gene-protein interactions using Ollivier-Ricci Curvature produces clusters of patients with defined prognostic significance, with high-risk groups harboring complex gene dysregulation impacting immune function.

The use of stereotactic body radiation therapy for ultracentral lung tumors is limited by increased toxicity. We hypothesized that using published normal tissue complication probability (NTCP) and tumor control probability (TCP) models could improve the therapeutic ratio between tumor control and toxicity. A proposed model-based approach was applied to virtually replan early-stage non-small cell lung cancer (NSCLC) tumors.The analysis included 63 patients with ultracentral NSCLC tumors treated at our center between 2008 and 2017. Along with current clinical constraints, additional NTCP model-based criteria, including for grade 3+ radiation pneumonitis (RP3+) and grade 2+ esophagitis, were implemented using 4 different fractionation schemes. Scaled dose distributions resulting in the highest TCP without violating constraints were selected (optimal plan [Planopt]). Planopt predictions were compared with the observed local control and toxicities.The observed 2-year local control rate was 72% (95% CI, 57%-88%) compared with 87% (range, 6%-93%) for Planopt TCP. Thirty-nine patients had Planopt with TCP > 80%, and 14 patients had Planopt TCP < 50%. The Planopt NTCPs for RP3+ were reduced by nearly half compared with patients' observed RP3+. The RP3+ NTCP was the most frequent reason for TCP of Planopt < 80% (14/24 patients), followed by grade 2+ esophagitis NTCP (5/24 patients) due to larger tumors (>40 cc vs ≤40 cc; P = .002) or a shorter tumor to esophagus distance (≥5 cm vs <5 cm; P < .001).We demonstrated the potential for model-based prescriptions to yield higher TCP while respecting NTCP for patients with ultracentral NSCLC. Individualizing treatments based on NTCP- and TCP-driven simulations halved the predicted relative to the observed rates of RP3+. Our simulations also identified patients whose TCP could not be improved without violating NTCP due to larger tumors or a near tumor to esophagus proximity.

The plasma cell cancer multiple myeloma (MM) varies significantly in genomic characteristics, response to therapy, and long-term prognosis. To investigate global interactions in MM, we combined a known protein interaction network with a large clinically annotated MM dataset. We hypothesized that an unbiased network analysis method based on large-scale similarities in gene expression, copy number aberration, and protein interactions may provide novel biological insights. Applying a novel measure of network robustness, Ollivier-Ricci Curvature, we examined patterns in the RNA-Seq gene expression and CNA data and how they relate to clinical outcomes. Hierarchical clustering using ORC differentiated high-risk subtypes with low progression free survival. Differential gene expression analysis defined 118 genes with significantly aberrant expression. These genes, while not previously associated with MM, were associated with DNA repair, apoptosis, and the immune system. Univariate analysis identified 8/118 to be prognostic genes; all associated with the immune system. A network topology analysis identified both hub and bridge genes which connect known genes of biological significance of MM. Taken together, gene interaction network analysis in MM uses a novel method of global assessment to demonstrate complex immune dysregulation associated with shorter survival.

Intensity modulated radiation therapy (IMRT) treatment planning is often formulated as the optimization of weights of fixed-geometry subfields (beamlets). Efficient optimization techniques can be based on direct storage of the influence matrix relating beamlet weights to dose values. However, direct storage of beamlet dose distributions for IMRT treatment planning can easily exceed several gigabytes, and is therefore often not feasible. We present a method for rapidly calculating full three-dimensional IMRT dose distributions, based on a vector of beamlet weights. The method is based on compressed beamlet dose distributions using fast digital wavelet transforms and so-called hard thresholding. We studied the method with a rectangular beamlet of cross section from a monoenergetic 6 MeV photon point source simulated in homogeneous (water) and heterogeneous (CT-data) phantoms. Dose was calculated using the accurate VMC++ Monte Carlo engine. The beamlet dose distributions were wavelet transformed and compressed by dropping wavelet coefficients below a given threshold value. Dose is then computed using the remaining wavelets. Selection of the wavelet basis function, decomposition level, and threshold values, for different slice orientations (transverse or parallel to the beam) and varying angles of beamlet incidence are studied. A typical in-slice compression ratio for a plane containing a beamlet was 32:1 using the sym2 wavelet and a threshold of 0.01, with a typical root-mean-square error, for voxels above 50% of the maximum dose, of about 0.04%. The overall compression performance, which includes many planes with little information content, is on the order of 100:1 or greater compared to full matrix storage. Although other methods are available to make the use of stored influence matrix values more feasible in IMRT treatment planning (such as using coarse grids or restricting values to defined volumes of interest) we conclude that wavelet compression facilitates the storage and use of full pencil dose deposition (influence matrix) data in IMRT treatment planning.

Since the early demonstration of the curative potential of radiation therapy for tumor sterilization, normal tissue toxicity continues to be dose limiting. Accurate prediction of patient¿s complication risk would allow personalization of treatment planning decisions. Nonlinear kernel methods can provide a robust framework for learning complex interactions between observed toxicities and treatment, anatomical, and patient-related variables. However, proper application of these powerful methods would require better understanding of a high-dimensional feature space that is spanned by all these variables. In this work, we investigate methods for visualization of this high-dimensional space and compare different approaches for extracting discriminant features. Our preliminary results demonstrate that principle component analysis is a valuable tool for visualizing high dimensional data and for determining proper kernel type. In addition, variable selection based on resampling methods within the logistic regression framework seemed to yield improved prediction performance compared to the recursive-feature elimination method.

We have attempted to better define host humoral immune response in neoplasia by quantitating serial circulating immune complex values before and after surgery in patients with primary or metastatic colorectal cancer. Circulating immune complex levels were correlated with serial carcinoembryonic antigen values and tumor courses in patients with primary resectable colorectal cancer (four patients), resectable liver metastases (three patients), diffuse liver metastases treated with regional chemotherapy (three patients), and untreated intrahepatic (one patient) and extrahepatic metastases (one patient). Circulating immune complex levels, as measured by an antigen-nonspecific assay, which utilized 4 percent polyethylene glycol insolubilization, were increased in all patients at presentation (734 delta OD450 +/- 381) when compared with normal human control sera (202 +/- 4, p less than 0.05). No particular relation was found between presenting circulating immune complex levels and tumor burden. Progressive circulating immune complex increases were demonstrated only in patients whose tumors were either completely removed or dramatically responded to regional therapy (that is, when the tumor antigen load, as reflected by the carcinoembryonic antigen value, rapidly diminished). Serum samples obtained at times of presumed antibody excess in the patients with gastrointestinal cancers were found to contain unexpectedly high concentrations of IgA. We believe these data demonstrate the kinetics of circulating immune complex change during tumor course and they have allowed us to begin to identify circulating immune complex components in patients with colorectal cancer. The results confirm our earlier findings in patients with gestational tumors and differ from accepted relations between immune complexes and tumor growth.

Purpose/Objective: To identify clinical, dosimetric, and location-related factors which correlate with local control in patients receiving definitive 3D conformal radiation therapy for non-small cell lung cancer (NSCLC). Materials/Methods: Between March 1991 and December 2001, 281 patients with NSCLC were treated with 3D-CRT to a dose ≥59 Gy and had archived treatment plans which could be recovered. Patients were excluded who had less than six months follow up in the absence of local/locoregional failure or any other failure (e.g. distant metastasis) without evidence of local/locoregional failure. 159 patients were evaluable. Patients with a discrete primary lesion exclusive of nodal regions (n = 56) were analyzed as a subset. Original treatment plans were imported into our research treatment planning system (CERR) for analysis. A representative PTV (1 cm expansion) was created around the original GTV. Dose and volume parameters for GTVs and PTVs were calculated. Tumor position was scaled from zero to one between the maximum extents of contoured lung tissue in three dimensions. Evaluated factors included clinical (age, gender, performance status, weight loss, smoking, histology, T stage, N stage), dosimetric (D80-D100, mean dose, maximum dose, and minimum dose), treatment factors (chemotherapy, treatment time, fraction size), as well as parameters derived from 3D-CRT planning information (tumor volume and GTV center position). A multivariate model was formed by logistic regression of linear combinations of candidate variables. To maximize the predictive power of our model extrapolated to other data sets, we first determined model complexity (number of parameters) and then estimated the most robust model. Model order was determined by using leave-one-out cross-validation, where all samples but one were used in round-robin testing. Bootstrap samples were re-fit to determine the most competitive models. Results: Median follow-up time for those patients still living was 32 months. Actuarial survival estimates (n=159) at two and four years were calculated for local tumor control (75% and 67%), cause specific survival (63% and 41%), and overall survival (53% and 30%). Patients receiving >66 Gy demonstrated improved local control at one year (86.8% vs. 59.7%, p = 0.017 using log-rank test, n=159). All multivariate modeling used the endpoint of local failure. Variables with highest univariate rank correlations (Rs) included GTV volume (0.21), GTV S-I position (0.165), N stage (0.158), T stage (0.149), PTV-D90 (0.144), and GTV-D100 (0.11). Leave-one-out analysis suggested an optimal model would include 6 parameters. Multiple 6-parameter models correlated with outcomes (highest Rs=0.336, p<0.00001). Primary tumors were then evaluated separately from nodal volumes. In the sub-set of patients with discrete primary tumors, strong univariate correlations (Rs) were seen with posterior tumor location (0.417), GTV volume (0.365), GTV-D95 (0.304), and GTV-D100 (0.292). Minimum distance from spinal cord to GTV surface correlated with these parameters, as well as with outcome (0.46). A multivariate model of frequently selected variables (age, GTV-D95, GTV-Volume, and GTV A-P tumor location) correlated strongly with local failure (Rs=0.604, p<0.00001). Conclusions: Local tumor control in isolated NSCLC primary tumors correlates strongly with parameters related to the high-dose portion of the DVH, clinical factors, and posterior location. Posterior tumor position strongly correlates with poor outcome.

In this paper, the current state of intensity-modulated radiation therapy (IMRT) treatment planning systems is reviewed, including some inefficiencies along with useful workarounds and potential advances. Common obstacles in IMRT treatment planning are discussed, including problems due to the lack of scatter tails in optimization dose calculations, unexpected hot spots appearing in uncontoured regions, and uncontrolled tradeoffs inherent in conventional systems. Workarounds that can be applied in current systems are reviewed, including the incorporation of an 'anchor zone' around the target volume (including a margin of separation), which typically induces adequate dose falloff around the target, and the use of pseudostructures to reduce conflicts among objective functions. We propose changing the planning problem statement so that different dosimetric or outcome goals are prioritized as part of the prescription ('prioritized prescription optimization'). Higher-priority goals are turned into constraints for iterations that consider lower-priority goals. This would control tradeoffs between dosimetric objectives. A plan review tool is proposed that specifically summarizes distances from a structure to hot or cold doses ('dose-distance plots'). An algorithm for including scatter in the optimization process is also discussed. Lastly, brief comments are made about the ongoing effort to use outcome models to rank or optimize treatment plans.

Abstract Drug-induced liver injury (DILI) is an adverse hepatic drug reaction that can potentially lead to life-threatening liver failure. Previously published work in the scientific literature on DILI has provided valuable insights for the understanding of hepatotoxicity as well as drug development. However, the manual search of scientific literature in PubMed is laborious. Natural language processing (NLP) techniques have been developed to decipher and understand the meaning of human language by extracting useful information from unstructured text data. In particular, NLP along with artificial intelligence (AI) / machine learning (ML) techniques may allow automatic processing of the DILI literature, but useful methods are yet to be demonstrated. To address this challenge, we have developed an integrated NLP/ML classification model to identify DILI-related literature using only paper titles and abstracts. We used 14,203 publications provided by the Critical Assessment of Massive Data Analysis (CAMDA) challenge, employing word vectorization techniques in NLP coupled with machine learning methods. Classification modeling was performed using 2/3 of the data for training and the remainder for testing in internal validation. The best performance was achieved using a linear support vector machine (SVM) model that combined vectors derived from term frequency-inverse document frequency (TF-IDF) and Word2Vec , achieving an accuracy of 95.0% and an F1-score of 95.0%. The final SVM model built using all 14,203 publications was tested on independent datasets, resulting in accuracies of 92.5%, 96.3%, and 98.3%, and F1-scores of 93.5%, 86.1%, and 75.6% for three test sets (T1-T3). The SVM model was tested on four external validation sets (V1-V4), resulting in accuracies of 92.0%, 96.2%, 98.3%, and 93.1%, and F1-scores of 92.4%, 82.9%, 75.0%, and 93.3%.

Lung tumors, especially those located close to or surrounded by soft tissues like the mediastinum, are difficult to segment due to the low soft tissue contrast on computed tomography images. Magnetic resonance images contain superior soft-tissue contrast information that can be leveraged if both modalities were available for training. Therefore, we developed a cross-modality educed learning approach where MR information that is educed from CT is used to hallucinate MRI and improve CT segmentation. Our approach, called cross-modality educed deep learning segmentation (CMEDL) combines CT and pseudo MR produced from CT by aligning their features to obtain segmentation on CT. Features computed in the last two layers of parallelly trained CT and MR segmentation networks are aligned. We implemented this approach on U-net and dense fully convolutional networks (dense-FCN). Our networks were trained on unrelated cohorts from open-source the Cancer Imaging Archive CT images (N=377), an internal archive T2-weighted MR (N=81), and evaluated using separate validation (N=304) and testing (N=333) CT-delineated tumors. Our approach using both networks were significantly more accurate (U-net $P <0.001$; denseFCN $P <0.001$) than CT-only networks and achieved an accuracy (Dice similarity coefficient) of 0.71$\pm$0.15 (U-net), 0.74$\pm$0.12 (denseFCN) on validation and 0.72$\pm$0.14 (U-net), 0.73$\pm$0.12 (denseFCN) on the testing sets. Our novel approach demonstrated that educing cross-modality information through learned priors enhances CT segmentation performance

Abstract In this note, we analyze two cancer dynamical models from a system-theoretic point of view. The first model is based upon stochastic controlled versions of the classical Lotka-Volterra equations. Here we consider from a controls point of view the utility of employing ultrahigh dose flashes in radiotherapy. The second is based on work of Norton-Simon-Massagué growth model that takes into account the heterogeneity of a tumor cell population. We indicate an optimal strategy based on linear quadratic control applied to a linear transformed model.

HN patients often experience multiple worsening symptoms during radiotherapy (RT), and some are unable to speak or have hoarseness. Physician driven systematic patient symptom questioning and obtaining comprehensible answers from the HN cancer patient can be challenging. To more efficiently monitor patient symptoms, a CTCAE v4.0 based hybrid patient and physician weekly questionnaire was created. The 18 question questionnaire used language adapted from the CTCAE for patient self-reporting. The patients completed the questionnaire while waiting for the weekly RT physician visit. The physician subsequently had the option of increasing the score of the nausea, anorexia, and/or aspiration question to grade 3 since these answers were solely based on clinical judgment. Symptoms were then managed at the discretion of the physician. The distributions of responses and weighted scores were calculated weekly, for 7 weeks of RT. The chi-square and Fisher's exact test were used to compare grade 2 or greater toxicity between weeks 1 and 7. A total of 58 patients receiving HN RT from May of 2010 to February of 2011 completed at least one questionnaire for a total of 352. The pain, 16 item, question was excluded from this analysis. Median age was 57.5 years. Tumor involvement included the oropharynx (52 %), larynx (26%), and oral cavity (17%). 52% of patients underwent definitive surgery, and 46% received chemotherapy. RT dose levels were 70 and 56 Gy (50 %), 66 and 54 Gy (28 %), and 60 and 52 Gy (17 %), respectively. Ordered from highest to lowest overall weighted severity at 7 weeks were: oral mucositis, voice changes, dysgeusia, dysphagia, anorexia, dry mouth, fatigue, trismus, constipation, nausea, vomiting, insomnia, dizziness, depression, anxiety, dyspepsia and aspiration. The last 6 adverse events remained relatively stable during treatment. Dry mouth, trismus, and nausea reached a plateau at week 3, 3 and 5, respectively. The remaining 8 adverse events increased throughout treatment despite symptom management interventions. A statistically significant difference in grade 2 or greater toxicity between weeks 1 and 7 was seen for voice changes (17 vs. 59 %, p < 0.01), oral mucositis (7 vs. 63%, p < 0.01), nausea (7 vs. 30%, p = 0.01), vomiting (2 vs. 18%, p = 0.04), dry mouth (12 vs. 50%, p < 0.01), dysgeusia (36 vs. 88%, p < 0.01), anorexia (12 vs. 59%, p < 0.01), and aspiration (2 vs. 18%, p = 0.04). The proposed hybrid CTCAE based questionnaire approach was feasible, and facilitated greater communication between the patient and physician. Improved interventions for HN RT symptoms are necessary.

Recent findings point toward the importance of cardiac substructures in thoracic radiotherapy. Incorporating an increased number of structures without adding considerable time to the clinical workflow call for automatic yet accurate segmentation. The aim of this work was to develop an atlas segmentation workflow including 14 cardiac substructures for contrast- and non-contrast-enhanced CT scans (CT+, CT-), and evaluate its performance relative to manual segmentations. One atlas was generated from manually segmented cardiac substructures in 20 CT+ scans, and another atlas for 20 CT- scans within a clinically available atlas software. The atlas performance was investigated in a subset of 10 scans/modality. In this subset, the modality-specific atlas was applied to each CT scan using a majority vote from the best three scans from the other 9 scans i.e. an overlap with at least two-thirds of the matching scans. The agreements between a reference observer and either the atlas or another observer were compared using the Dice Similarity Coefficient (DSC; significance: p≤0.05 using the Wilcoxon signed-rank test), and volumetric trends were assessed as the normalized relative volume difference (RVD). Ultimately, the time required to apply the atlas was compared to that of the manual segmentation. For both modalities, the atlas performance was overall satisfying for 9/14 structures: the aorta, chambers, heart, inferior and superior vena cava, and pulmonary artery, but the DSCs between the atlas and the reference observer were significantly lower (p<0.0001) vs. the DSCs between the reference observer and the other observer: structure averaged DSCs=0.52-0.91 vs. 0.86-0.97 for CT+; DSCs=0.39-0.90 vs. 0.70-0.97 for CT-. The atlas systematically over- and underestimated the reference observer for CT+ and CT-, respectively (RVD=0.02-0.40, and -0.02--0.73). Regardless of structure and modality, deviations between the atlas and the reference observer were typically observed in the coronary vessels, as well as in all boundary slices. On average, the atlas workflow required 2-3 min/scan and post-processing, including also the coronary vessels, added 15 min/scan in five scans for each modality. The average time required for manual segmentation was 1.5 h/scan. Our findings suggest that atlas segmentation is suitable for the nine major cardiac substructures for both contrast and non-contrast-enhanced CT scans. The time required to apply and post-process the atlas segmentations was considerably shorter compared to that of the manual segmentation (67% time saved), and the atlas could, therefore, also be used for the coronary vessels as an initial segmentation to minimize added load to the clinical workflow.

Purpose: Recently, radiomics has emerged as a new research field with the aim of (1) identifying quantitative medical image features associated with outcomes, (2) building predictive models of outcomes, (3) and thereby better understanding the underlying mechanisms of outcomes. Our group has developed methods to extract quantitative features from medical images and to model the outcomes. In this work, we summarize our studies with magnetic resonance (MR) images, demonstrating the potential of using radiomics for outcomes research. Methods: This retrospective study analyzed 178 women with invasive ductal carcinoma (IDC) and preoperative breast MR images. Tumor subtypes defined by immunohistochemistry surrogates are: estrogen and progesterone receptor positive (ERPR+; n=95), HER2 receptor positive (HER2+; n=35) and triple negative (TN; n=48). Tumors were contoured on a single central slice from fat-suppressed T1-weight pre- and three post-contrast images. Image features were extracted from the contours. Clinical and pathologic features were collected. Linear regression analysis was used to build a predictive model of the FDA approved OncotypeDx Breast Cancer 21-gene Assay Recurrence Score (RS) that was measured for ER positive patients. In addition, a machine learning method was used to differentiate three breast cancer subtypes using image features. Results: Using stepwise multiple linear regression analysis, a model with three features to predict the OncotypeDx RS achieved R-squared = 0.228 (adjusted R-squared = 0.198; p = 0.0002) and Spearman's rank correlation coefficient = 0.485 (p < 0.0001). Separately, a support vector machine model with 9 features distinguished IDC subtypes with an overall accuracy of 83.4%: 89.2% (ERPR+), 63.6% (HER2+) and 82.5% (TN). A Kruskal-Wallis test for the 9 features showed a statistically significant difference between ERPR+, HER2+ and TN subtypes with p < 0.0001. Conclusion: In this study, we demonstrated the potential to use radiomics to understand breast cancer genomics with non-invasive MR images.

Abstract In this work, we introduce a new mathematical framework based on network curvature to extract significant cancer subtypes from multi-omics data. This extends our previous work that was based on analyzing a fixed single-omics data class (e,g, CNA, gene expression, etc.). Notably, we are able to show that this new methodology provided us with significant survival differences on Kaplan-Meier curves across almost every cancer that we considered. Moreover, the variances in Ollivier-Ricci curvature was explored to investigate its usefulness in network topology analysis as this curvature may be capturing subtle functional changes between various cancer subtypes.

Radiation exposure of healthy cells can halt cell cycle temporarily or permanently. In this work, we analyze the time evolution of p21 and p53 from two single cell datasets of retinal pigment epithelial cells exposed to several levels of radiation, and in particular, the effect of radiation on cell cycle arrest. Employing various quantification methods from signal processing, we show how p21 levels, and to a lesser extent p53 levels, dictate whether the cells are arrested in their cell cycle and how frequently these mitosis events are likely to occur. We observed that single cells exposed to the same dose of DNA damage exhibit heterogeneity in cellular outcomes and that the frequency of cell division is a more accurate monitor of cell damage rather than just radiation level. Finally, we show how heterogeneity in DNA damage signaling is manifested early in the response to radiation exposure level and has potential to predict long-term fate.

To predict the risk of radiation pneumonitis/fibrosis (grade 2 or greater) following definitive non-small cell lung cancer treatment based on clinical and dosimetric parameters including spatial dose positioning and pre-treatment pulmonary function tests. Two hundred (200) patients were analyzed, including all patients with recoverable plan archives and greater than 6 mos. follow-up treated using 3-D treatment planning between June 1991 and December 2001. There were 27 patients with grade 2, 16 with grade 3, 1 with grade 4, and 0 grade 5 complications. The treatment plans were processed using our research treatment planning environment (CERR). CERR provides programmable access to all the treatment planning data including dose distributions and anatomic structures. Derived dosimetric parameters included total lung volume (including the GTV), GTV volume, cumulative dose-volume histogram parameters Vx, in 5 Gy steps beginning at 5 Gy, mean normal lung dose, maximum lung dose, mean GTV dose, and the generalized equivalent uniform dose (GEUD) for the total normal lungs with an exponential parameter chosen to univariately maximize the correlation with complications. Treatment planning information was supplemented with patient-specific clinical parameters including neoadjuvant chemotherapy, concurrent chemotherapy, pulmonary function tests (pre-treatment FEV1), smoking history, age, gender, KPS, and pre-treatment weight loss. Multivariate logistic regression was used to build multi-variable models. Bootstrap techniques were used to test the stability of the regression variable selection and. Goodness of fit was determined using the F-test. Wald’s statistics were used to check the significance of individual variables in the logistic model. Spearman’s rank correlation coefficient (Rs) was used to measure model correlation with complications. To quantify the effect of the high-dose location within the lungs, we used CERR to compute the center of the GTV (denoted GTV-SI) relative to the maximum inferior and superior extent of the lung. Similarly, the GTV location was quantified anterior-posterior and left-right. The best GEUD exponential parameter was determined univariately to be either 5.31 (for heterogeneity-corrected plans) or 5.58 (for water-based plans), representing a weighted high dose ( >40 Gy) average. This parameter was then fixed for multivariate logistic regression. The effect on the resulting model of analyzing water-based plans vs. heterogeneity-corrected plans (ratio-TAR method) was generally modest (e.g., Rs improved from 0.29 (water-based) to 0.32 (heterogeneity-corrected)). Using the water-based plans (n = 200), the best model (Rs = 0.249; p < 0.001) included (in order of significance): (1) concurrent chemotherapy, (2) GEUD, (3) V60 (with a negative coefficient), (4) GTV-SI (base irradiation increases risk), (5) V80 (with a negative coefficient), and (6) V5 (with a positive coefficient). V60 and V80 can be thought of as modifying the behavior of GEUD. However, the effect of V5 did not reach significance according to Wald’s test. Similar analyses based on subsets of patients with heterogeneous dose calculations and including data from pre-treatment pulmonary function tests resulted in the same four most significant factors. A multi-variable logistic-regression model was significantly correlated with pneumonitis/fibrosis in our patient cohort and is therefore expected to predict pneumonitis/fibrosis risk. The parameters which were consistently correlated with high risk are: (1) concurrent chemotherapy, (2) irradiation of large volumes to high doses (represented by the combination of GEUD and V60), and (3) tumor location in the base of either lung

Recent years have witnessed tremendous progress in IGRT technology and potential possibilities for adapting treatment planning on a daily basis. However, there is a lack in having software tools toward this goal. Therefore, we have implemented a software suite, DIRART, for deformable image registration (DIR) plus adaptive radiotherapy (ART) research. DIRART is a large set of programs developed using MATLAB. It contains DIR algorithms, common ART functions, integrated graphics user interfaces, visualization features, and dose metrics analysis functions. In addition, it complementarily works together with CERR (Computational Environment for Radiotherapy Research) to offer more functions. DIRART is designed around the concepts of the interactive RT objects, including images, structures, doses and deformation vector fields (DVF) etc. By exchanging RT objects with TPS via DICOM-RT files, DIRART provides a full featured working environment for ART related research tasks. It is capable of transforming dose distributions and structures between the planning CT and the daily images according to the computed DVF so that such RT information could be viewed and evaluated on either image sets. It can also rescale, subtract and sum up the transformed doses, and convert isodose lines to structures. For DIR applications, DIRART is a toolbox which provides 20+ DIR algorithms, including the newer inverse consistency algorithms to provide consistency DVF in both directions with better accuracy. As a DIR research environment, DIRART dose not only provide a good set of image preprocessing algorithms and post-processing functions to facilitate the development and testing of DIR algorithms, but also offers a good amount of options for results visualization, evaluation and validation. DIRART is designed in a data-oriented style with focus on usability, user-friendliness, performance, accuracy, flexibility, features, configurability and stability. It has great potential for the ART and DIR research.

To investigate if local morphological variations in FDG-PET uptake could be used to predict a priori local tumor control from pre-treatment diagnostic scanning and onset of lung injury from post-treatment scanning. We are investigating new methods for analyzing PET uptake features in different cancer sites. Besides methods based on statistical descriptors of SUV and tracer kinetics, we are exploring local image features as potential prognostic factors for assessing outcomes. One approach is based on intensity-volume histograms (IVH). The IVH method enables metrics to be extracted from PET images such as Ix (minimum intensity to x% highest intensity volume) and Vx (% volume having at least x% intensity value). The other approach involves extraction of morphological features from regions of interest (ROI) such as shape deformations and texture homogeneity. To explore this methodology we analyzed 17 NSCLC patients who received 3D-CRT with a mean F/U of 1 year. 16 patients with pre-RT FDG-PET were analyzed for local failure (37.5% rate) with gross tumor volume (GTV) as the ROI. 9 patients with post-RT FDG-PET were analyzed for pneumonitis (22.2% rate) with lung minus GTV as the ROI. About 30 relevant variables were extracted from each case, which included: ROI volume, SUV descriptors, total lesion glycolosis (TLG), Ix, Vx, and local texture variability metrics. Model building approaches based on logistic regression and machine learning were evaluated. Statistical association was measured using Spearman's rank correlation (rs). In local failure analysis, GTV and TLG had the highest correlation (rs = 0.476, p = 0.031) followed by I90 (rs = 0.434, p = 0.046) and V90 (rs = -0.420, p = 0.053), while meanSUV (rs = 0.252, p = 0.173), maxSUV (rs = 0.364, p = 0.083), and local texture homogeneity (rs = 0.196, p = 0.233) showed modest association. A combined logistic model of TLG and V90 yielded rs = 0.616 (p = 0.009). This is slightly improved using a quadratic kernel mapping (rs = 0.644, p = 0.006). In pneumonitis analysis, local contrast (rs = 0.725, p = 0.014) and local homogeneity (rs = −0.725, p = 0.014) had the highest correlation in agreement with previous findings (Hicks et al., IJROBP '04). Other noticeable correlates included V90 (rs = 0.621, p = 0.037), Lung minus GTV volume (rs = -0.518, p = 0.077), and maxSUV (rs = −0.416, p = 0.133). We have explored new approaches based on pattern recognition analysis of FDG-PET uptake to predict outcomes in lung cancer patients post-RT treatment. In our preliminary analysis, we observed strong intensity-volume effect in predicting local failure and significant local heterogeneity association with onset of pneumonitis. Analyses on larger datasets would be required to elucidate these observations.

Purpose: We previously developed a novel approach to tumor response modeling which was inspired by the gEUD model of tumor cell kill under inhomogeneous dose profiles. The model assumes that the cell‐kill law is valid, but further adds a parameter analogous to the generalized equivalent uniform dose model to test the independence of response of tumor subvolumes. The purpose of this study was to test the resulting generalized tumor dose (gTD) model on an independent lung cancer subset. Methods: Niemierko's gEUD model of cell kill introduces a generalization term describing interactions among tumor sub‐volumes under non‐uniform dose distributions. This parameter effectively applies weights to sub‐volumes based off of tumor type‐specific sensitivities to variance in dose. We combine the gEUD with the cell‐kill based cEUD to form the “generalized tumor dose” (gTD) model, which is given by gTD=‐α −1 *ln(Vt/Vref)‐ (α*a) −1 *ln(Σi=1:subvolsM#subvols −1 *e (‐α*a*di)). This model was previously applied to lung and H&amp;N data. Here, we apply the model to an independent lung dataset (N= 36). Results: We found that the model does not perform well for large tumors (volume &gt; 70 cc.) However,. However, for tumors smaller than this volume, the model was well‐correlated with local control, with a normalized log‐likelihood value (log‐likelihood divided by the number of cases) of −0.604. Conclusion: The presented model continues to predict tumor response in this dataset, for tumors less than 70 cc. To date, no other TCP model has been validated on clinical datasets for tumors with varying volumes and dose distributions. More tests are needed to further refine and validate the model.

You have accessJournal of UrologyProstate Cancer: Localized I1 Apr 2010285 OVERALL AND CANCER SPECIFIC SURVIVAL FOLLOWING DEFINITIVE THERAPY FOR CLINICALLY LOCALIZED PROSTATE CANCER IN THE PROSTATE-SPECIFIC ANTIGEN ERA Andrew Stephenson, Eric Klein, Jay Ciezki, Chandana Reddy, Michael Kattan, Changhong Yu, Joseph Deasy, Jeff Michalski, Donna Kallogjeri, Jessica Lubahn, Jason Luly, Jay Piccirillo, and Adam Kibel Andrew StephensonAndrew Stephenson Cleveland, OH More articles by this author , Eric KleinEric Klein Cleveland, OH More articles by this author , Jay CiezkiJay Ciezki Cleveland, OH More articles by this author , Chandana ReddyChandana Reddy Cleveland, OH More articles by this author , Michael KattanMichael Kattan Cleveland, OH More articles by this author , Changhong YuChanghong Yu Cleveland, OH More articles by this author , Joseph DeasyJoseph Deasy St. Louis, MO More articles by this author , Jeff MichalskiJeff Michalski St. Louis, MO More articles by this author , Donna KallogjeriDonna Kallogjeri St. Louis, MO More articles by this author , Jessica LubahnJessica Lubahn St. Louis, MO More articles by this author , Jason LulyJason Luly St. Louis, MO More articles by this author , Jay PiccirilloJay Piccirillo St. Louis, MO More articles by this author , and Adam KibelAdam Kibel St. Louis, MO More articles by this author View All Author Informationhttps://doi.org/10.1016/j.juro.2010.02.347AboutPDF ToolsAdd to favoritesDownload CitationsTrack CitationsPermissionsReprints ShareFacebookTwitterLinked InEmail INTRODUCTION AND OBJECTIVES Radical prostatectomy, external-beam radiotherapy (EBRT), and brachytherapy are accepted treatment options for clinically localized prostate cancer. Comparative survival data from randomized trials is lacking. We endeavored to determine if survival differences exist between these treatments in a cohort of patients treated according to contemporary surgical and radiation therapy treatment standards. METHODS Between 1995 and 2005, 10,472 patients with localized prostate cancer were treated by radical prostatectomy (N=6493), EBRT (N=2260), and brachytherapy (N=1719) at Cleveland Clinic and Barnes-Jewish Hospital. Comorbidity was assessed by a chart-based review. Cox proportional hazards regression analysis and Fine and Gray competing risk regression analysis was used to model the disease- and patient-specific parameters for overall survival and cancer-specific survival, respectively. Propensity score analysis was used to adjust for differences in observed background characteristics. RESULTS Using propensity score analysis, EBRT and brachytherapy were significantly associated with diminished survival (HR 1.6 [95% CI: 1.4-1.9] and 1.7 [95% CI: 1.4-2.1], respectively; P < 0.001) compared to radical prostatectomy after adjusting for biopsy Gleason score, PSA, age, comorbidity, ethnicity, and clinical stage (all P ¡Ü 0.002). EBRT and brachytherapy were also associated with a significantly higher rate of androgen-deprivation therapy (P < 0.001). Compared to radical prostatectomy, a trend towards higher cancer-specific mortality was observed for EBRT (adjusted HR 1.6; 95% CI: 1.0-2.6), but not for brachytherapy (adjusted HR 1.1; 95% CI: 0.5, 2.6). CONCLUSIONS After adjusting for the most relevant disease-specific and patient-specific confounders, radical prostatectomy is associated with improved intermediate-term survival compared to EBRT and brachytherapy. Physicians and patients should consider these potential survival differences when choosing among treatment options for localized prostate cancer. © 2010 by American Urological Association Education and Research, Inc.FiguresReferencesRelatedDetails Volume 183Issue 4SApril 2010Page: e113 Advertisement Copyright & Permissions© 2010 by American Urological Association Education and Research, Inc.MetricsAuthor Information Andrew Stephenson Cleveland, OH More articles by this author Eric Klein Cleveland, OH More articles by this author Jay Ciezki Cleveland, OH More articles by this author Chandana Reddy Cleveland, OH More articles by this author Michael Kattan Cleveland, OH More articles by this author Changhong Yu Cleveland, OH More articles by this author Joseph Deasy St. Louis, MO More articles by this author Jeff Michalski St. Louis, MO More articles by this author Donna Kallogjeri St. Louis, MO More articles by this author Jessica Lubahn St. Louis, MO More articles by this author Jason Luly St. Louis, MO More articles by this author Jay Piccirillo St. Louis, MO More articles by this author Adam Kibel St. Louis, MO More articles by this author Expand All Advertisement Advertisement PDF downloadLoading ...

Abstract Purpose Delineating the swallowing and chewing structures in Head and Neck (H&amp;N) CT scans is necessary for radiotherapy treatment (RT) planning to reduce the incidence of radiation-induced dysphagia, trismus, and speech dysfunction. Automating this process would decrease the manual input required and yield reproducible segmentations, but generating accurate segmentations is challenging due to the complex morphology of swallowing and chewing structures and limited soft tissue contrast in CT images. Methods We trained deep learning models using 194 H&amp;N CT scans from our institution to segment the masseters (left and right), medial pterygoids (left and right), larynx, and pharyngeal constrictor muscle using DeepLabV3+ with the resnet-101 backbone. Models were trained in a sequential manner to guide the localization of each structure group based on prior segmentations. Additionally, an ensemble of models was developed using contextual information from three different views (axial, coronal, and sagittal), for robustness to occasional failures of the individual models. Output probability maps were averaged, and voxels were assigned labels corresponding to the class with the highest combined probability. Results The median dice similarity coefficients (DSC) computed on a hold-out set of 24 CT scans were 0.87±0.02 for the masseters, 0.80±0.03 for the medial pterygoids, 0.81±0.04 for the larynx, and 0.69±0.07for the constrictor muscle. The corresponding 95 th percentile Hausdorff distances were 0.32±0.08cm (masseters), 0.42±0.2cm (medial pterygoids), 0.53±0.3cm (larynx), and 0.36±0.15cm (constrictor muscle). Dose-volume histogram (DVH) metrics previously found to correlate with each toxicity were extracted from manual and auto-generated contours and compared between the two sets of contours to assess clinical utility. Differences in DVH metrics were not found to be statistically significant (p&gt;0.05) for any of the structures. Further, inter-observer variability in contouring was studied in 10 CT scans. Automated segmentations were found to agree better with each of the observers as compared to inter-observer agreement, measured in terms of DSC. Conclusions We developed deep learning-based auto-segmentation models for swallowing and chewing structures in CT. The resulting segmentations can be included in treatment planning to limit complications following RT for H&amp;N cancer. The segmentation models developed in this work are distributed for research use through the open-source platform CERR, accessible at https://github.com/cerr/CERR .

Although deep convolutional networks have been widely studied for head and neck (HN) organs at risk (OAR) segmentation, their use for routine clinical treatment planning is limited by a lack of robustness to imaging artifacts, low soft tissue contrast on CT, and the presence of abnormal anatomy. In order to address these challenges, we developed a computationally efficient nested block self-attention (NBSA) method that can be combined with any convolutional network. Our method achieves computational efficiency by performing non-local calculations within memory blocks of fixed spatial extent. Contextual dependencies are captured by passing information in a raster scan order between blocks, as well as through a second attention layer that causes bi-directional attention flow. We implemented our approach on three different networks to demonstrate feasibility. Following training using 200 cases, we performed comprehensive evaluations using conventional and clinical metrics on a separate set of 172 test scans sourced from external and internal institution datasets without any exclusion criteria. NBSA required a similar number of computations (15.7 gflops) as the most efficient criss-cross attention (CCA) method and generated significantly more accurate segmentations for brain stem (Dice of 0.89 vs. 0.86) and parotid glands (0.86 vs. 0.84) than CCA. NBSA's segmentations were less variable than multiple 3D methods, including for small organs with low soft-tissue contrast such as the submandibular glands (surface Dice of 0.90).

Purpose: evaluate a commercial synthetic CT (syn-CT) software for use in prostate radiotherapy Methods: Twenty prostate patients underwent CT and MR simulation scans in treatment position on a 3T Philips scanner. The MR protocol consisted of a T2w turbo spin-echo for soft tissue contrast, a 2D balanced-fast field echo (b-FFE) for fiducial identification, a dual-echo 3D FFE B0 map for distortion analysis and a 3D mDIXON FFE sequence to generate syn-CT. Two echoes are acquired during mDIXON scan, allowing water, fat, and in-phase images to be derived using the frequency shift of the fat and water protons. Tissues were classified as: air, adipose, water, trabecular/spongy bone and compact/cortical bone and assigned specific bulk HU values. Bone structures are segmented based on a pelvis bone atlas. Accuracy of syn-CT for patient treatment planning was analyzed by transferring the original plan and structures from the CT to syn-CT via rigid registration and recalculating dose. In addition, new IMRT plans were generated on the syn-CT using structures contoured on MR and transferred to the syn-CT. Accuracy of fiducial-based localization at the treatment machine performed using syn-CT or DRRs generated from syn-CT was assessed by comparing to orthogonal kV radiographs or CBCT. Results: Dosimetric comparison between CT and syn-CT was within 0.5% for all structures. The de-novo optimized plans generated on the syn-CT met our institutional clinical objectives for target and normal structures. Patient-induced susceptibility distortion based on B0 maps was within 1mm and 0.4 mm in the body and prostate. The rectal and bladder outlines on the syn-CT were deemed sufficient for assessing rectal and bladder filling on the CBCT at the time of treatment. CBCT localization showed a median error of < ±1 mm in LR, AP and SI direction. Conclusion: MRI derived syn-CT can be used clinically in MR-alone planning and treatment process for prostate. Drs. Deasy, Hunt and Tyagi have Master research agreement with Philips healthcare.

In radiotherapy treatment planning of cancer patients, the collection of multiple images of different and yet complementary information is rapidly becoming the norm. Beside CT data sets, PET and/or MRI or MRS images are also being used to aid in the definition of the target volume for treatment optimization. We are investigating methods to integrate available information for joint target registration and segmentation of multi-modality images as perceived by the human observer. Towards this goal, we are exploring multi-valued level set deformable models in conjunction with human perception models for simultaneous delineation of multi-modality images consisting of combinations of PET, CT, or MR datasets. Information from multimodality image sets is integrated based on a logical model to define the final target volume. The methods were demonstrated qualitatively on patient cases of lung cancer with PET/CT and a prostate patient case with CT and MR. We used a series of phantom data of CT, PET, and MR for quantification analysis. Phantom studies suggest 90% segmentation accuracy and less than 2% volume error when integrating all of the three modalities. This is compared with 74% accuracy and 4.4% volume error when using CT-based systems. These results indicate that this semi-automated multimodality-based definition of the biophysical target would provide a feasible and accurate framework for integrating complementary imaging information from different modalities and potentially a useful tool for optimizing of cancer patients radiotherapy plans.

Purpose: To generate automatic segmentation of head and neck normal structures from multi-parametric MR Dixon images. Materials and Method: We present a multi-atlas based registration combined with machine learning-based segmentation of head and neck structures from MR T1 FFE based mdixon images. All the images were acquired using a 3T Phillips MR scanner. 6 patients; 3 in atlas and 3 in testing were used. The individual mdixon images were registered to corresponding images from the multi-atlas using affine and B-spline deformable registration using the open-source software Plastimatch. Second, the best-aligned image pairs were automatically extracted through automatic landmark generation and matching. Scale Invariant Feature Transform (SIFT) features were used to generate the landmarks. The segmentation labels were propagated from the best matching atlas. A random forest (RF) classifier trained with K=10 fold cross validation using the MR mdixon and Haralick textures computed on the same images from the multi-atlas refined the segmentation. Finally, the generated segmentations were smoothed using Markov Random Field and morphological post-processing. Results: The patients used in the analysis displayed anatomical variations owing to dental implants and disease. We evaluated the segmentations generated by our method by computing dice overlap scores with manually generated segmentations. Our method resulted in an accuracy ranging between 0.5 to 0.73 for the various structures, namely, bone, right and left parotid, right and left submandibular glands. The algorithm selected registrations closely agreed with the visual comparison of the image registrations. Conclusions: We developed a fully automatic method for normal structures segmentation that combines multi-atlas based registration with machine learning from multi parametric MRI. Our method quantifies the accuracy of registrations by using automatic landmark extraction. Accurate, automatic volumetric segmentation of normal structures is essential for MR-based treatment planning.

&lt;div&gt;Abstract&lt;p&gt;Tissue survival responses to ionizing radiation are nonlinear with dose, rather yielding tissue-specific descending curves that impede straightforward analysis of biologic effects. Apoptotic cell death often occurs at low doses, while at clinically relevant intermediate doses, double-strand break misrepair yields mitotic death that determines outcome. As researchers frequently use a single low dose for experimentation, such strategies may inaccurately depict inherent tissue responses. Cutting edge radiobiology has adopted full dose survival profiling and devised mathematical algorithms to fit curves to observed data to generate highly reproducible numerical data that accurately define clinically relevant inherent radiosensitivities. Here, we established a protocol for irradiating organoids that delivers radiation profiles simulating the organ of origin. This technique yielded highly similar dose–survival curves of small and large intestinal crypts &lt;i&gt;in vivo&lt;/i&gt; and their cognate organoids analyzed by the single-hit multi-target (SHMT) algorithm, outcomes reflecting the inherent radiation profile of their respective Lgr5&lt;sup&gt;+&lt;/sup&gt; stem cell populations. As this technological advance is quantitative, it will be useful for accurate evaluation of intestinal (patho)physiology and drug screening.&lt;/p&gt;Significance:&lt;p&gt;These findings establish standards for irradiating organoids that deliver radiation profiles that phenocopy the organ of origin.&lt;/p&gt;&lt;p&gt;&lt;i&gt;See related commentary by Muschel et al., p. 927&lt;/i&gt;&lt;/p&gt;&lt;/div&gt;

&lt;p&gt;Supplementary Materials. Table S1. List of all reagents used in this study; Figure S1. LI organoids regrow after radiation; Figure S2. Sorting strategy for intestinal stem cells from the small and large intestines; Figure S3. Flow cytometry analysis of GFP+ cells in organoids versus colonies; Figure S4. Day 1 post-plating colonic stem cell colonies display defective DNA repair.&lt;/p&gt;

&lt;p&gt;Supplementary Materials. Table S1. List of all reagents used in this study; Figure S1. LI organoids regrow after radiation; Figure S2. Sorting strategy for intestinal stem cells from the small and large intestines; Figure S3. Flow cytometry analysis of GFP+ cells in organoids versus colonies; Figure S4. Day 1 post-plating colonic stem cell colonies display defective DNA repair.&lt;/p&gt;

Supplementary Table 1 from A MicroRNA Expression Signature for Cervical Cancer Prognosis

&lt;p&gt;The study analysis plan, Figures S1-S6, Table S1&lt;/p&gt;

&lt;p&gt;Table S1. Excluded cohorts in the analysis and the reason for exclusion; Table S2. The datasets used for this analysis in three different groups based on Mehta, et al.'s data (2012) and the estimated EQD22.8 (Gy) of the best-fit by the model simulation; Table S3. The additional datasets used for external validation of the model analysis; Figure S1. Deviations of estimated EQD210,model from BED; Figure S2. Dose-response curves estimated from maximum likelihood method based on the EQD210,model values; Figure S4. Dose-response curves obtained by applying other models to the dataset of current study: (A) Shuryak et al.'s model (69); and (B) Tai et al.'s model (70).&lt;/p&gt;

&lt;p&gt;The study analysis plan, Figures S1-S6, Table S1&lt;/p&gt;

&lt;div&gt;AbstractPurpose:&lt;p&gt;To quantitatively predict the impact of cardiopulmonary dose on overall survival (OS) after radiotherapy for locally advanced non–small cell lung cancer.&lt;/p&gt;Experimental Design:&lt;p&gt;We used the NRG Oncology/RTOG 0617 dataset. The model building procedure was preregistered on a public website. Patients were split between a training and a set-aside validation subset (&lt;i&gt;N&lt;/i&gt; = 306/131). The 191 candidate variables covered disease, patient, treatment, and dose-volume characteristics from multiple cardiopulmonary substructures (atria, lung, pericardium, and ventricles), including the minimum dose to the hottest x% volume (Dx%[Gy]), mean dose of the hottest x% (MOHx%[Gy]), and minimum, mean (Mean[Gy]), and maximum dose. The model building was based on Cox regression and given 191 candidate variables; a Bonferroni-corrected &lt;i&gt;P&lt;/i&gt; value threshold of 0.0003 was used to identify predictors. To reduce overreliance on the most highly correlated variables, stepwise multivariable analysis (MVA) was repeated on 1000 bootstrapped replicates. Multivariate sets selected in ≥10% of replicates were fit to the training subset and then averaged to generate a final model. In the validation subset, discrimination was assessed using Harrell &lt;i&gt;c&lt;/i&gt;-index, and calibration was tested using risk group stratification.&lt;/p&gt;Results:&lt;p&gt;Four MVA models were identified on bootstrap. The averaged model included atria D45%[Gy], lung Mean[Gy], pericardium MOH55%[Gy], and ventricles MOH5%[Gy]. This model had excellent performance predicting OS in the validation subset (&lt;i&gt;c&lt;/i&gt; = 0.89).&lt;/p&gt;Conclusions:&lt;p&gt;The risk of death due to cardiopulmonary irradiation was accurately modeled, as demonstrated by predictions on the validation subset, and provides guidance on the delivery of safe thoracic radiotherapy.&lt;/p&gt;&lt;/div&gt;

&lt;div&gt;AbstractPurpose:&lt;p&gt;To quantitatively predict the impact of cardiopulmonary dose on overall survival (OS) after radiotherapy for locally advanced non–small cell lung cancer.&lt;/p&gt;Experimental Design:&lt;p&gt;We used the NRG Oncology/RTOG 0617 dataset. The model building procedure was preregistered on a public website. Patients were split between a training and a set-aside validation subset (&lt;i&gt;N&lt;/i&gt; = 306/131). The 191 candidate variables covered disease, patient, treatment, and dose-volume characteristics from multiple cardiopulmonary substructures (atria, lung, pericardium, and ventricles), including the minimum dose to the hottest x% volume (Dx%[Gy]), mean dose of the hottest x% (MOHx%[Gy]), and minimum, mean (Mean[Gy]), and maximum dose. The model building was based on Cox regression and given 191 candidate variables; a Bonferroni-corrected &lt;i&gt;P&lt;/i&gt; value threshold of 0.0003 was used to identify predictors. To reduce overreliance on the most highly correlated variables, stepwise multivariable analysis (MVA) was repeated on 1000 bootstrapped replicates. Multivariate sets selected in ≥10% of replicates were fit to the training subset and then averaged to generate a final model. In the validation subset, discrimination was assessed using Harrell &lt;i&gt;c&lt;/i&gt;-index, and calibration was tested using risk group stratification.&lt;/p&gt;Results:&lt;p&gt;Four MVA models were identified on bootstrap. The averaged model included atria D45%[Gy], lung Mean[Gy], pericardium MOH55%[Gy], and ventricles MOH5%[Gy]. This model had excellent performance predicting OS in the validation subset (&lt;i&gt;c&lt;/i&gt; = 0.89).&lt;/p&gt;Conclusions:&lt;p&gt;The risk of death due to cardiopulmonary irradiation was accurately modeled, as demonstrated by predictions on the validation subset, and provides guidance on the delivery of safe thoracic radiotherapy.&lt;/p&gt;&lt;/div&gt;

&lt;div&gt;Abstract&lt;p&gt;&lt;b&gt;Purpose:&lt;/b&gt; To demonstrate that a mathematical model can be used to quantitatively understand tumor cellular dynamics during a course of radiotherapy and to predict the likelihood of local control as a function of dose and treatment fractions.&lt;/p&gt;&lt;p&gt;&lt;b&gt;Experimental Design:&lt;/b&gt; We model outcomes for early-stage, localized non–small cell lung cancer (NSCLC), by fitting a mechanistic, cellular dynamics-based tumor control probability that assumes a constant local supply of oxygen and glucose. In addition to standard radiobiological effects such as repair of sub-lethal damage and the impact of hypoxia, we also accounted for proliferation as well as radiosensitivity variability within the cell cycle. We applied the model to 36 published and two unpublished early-stage patient cohorts, totaling 2,701 patients.&lt;/p&gt;&lt;p&gt;&lt;b&gt;Results:&lt;/b&gt; Precise likelihood best-fit values were derived for the radiobiological parameters: α [0.305 Gy&lt;sup&gt;−1&lt;/sup&gt;; 95% confidence interval (CI), 0.120–0.365], the α/β ratio (2.80 Gy; 95% CI, 0.40–4.40), and the oxygen enhancement ratio (OER) value for intermediately hypoxic cells receiving glucose but not oxygen (1.70; 95% CI, 1.55–2.25). All fractionation groups are well fitted by a single dose–response curve with a high &lt;i&gt;χ&lt;sup&gt;2&lt;/sup&gt; P&lt;/i&gt; value, indicating consistency with the fitted model. The analysis was further validated with an additional 23 patient cohorts (&lt;i&gt;n&lt;/i&gt; = 1,628). The model indicates that hypofractionation regimens overcome hypoxia (and cell-cycle radiosensitivity variations) by the sheer impact of high doses per fraction, whereas lower dose-per-fraction regimens allow for reoxygenation and corresponding sensitization, but lose effectiveness for prolonged treatments due to proliferation.&lt;/p&gt;&lt;p&gt;&lt;b&gt;Conclusions:&lt;/b&gt; This proposed mechanistic tumor-response model can accurately predict overtreatment or undertreatment for various treatment regimens. &lt;i&gt;Clin Cancer Res; 23(18); 5469–79. ©2017 AACR&lt;/i&gt;.&lt;/p&gt;&lt;/div&gt;

&lt;div&gt;Abstract&lt;p&gt;&lt;b&gt;Purpose:&lt;/b&gt; To demonstrate that a mathematical model can be used to quantitatively understand tumor cellular dynamics during a course of radiotherapy and to predict the likelihood of local control as a function of dose and treatment fractions.&lt;/p&gt;&lt;p&gt;&lt;b&gt;Experimental Design:&lt;/b&gt; We model outcomes for early-stage, localized non–small cell lung cancer (NSCLC), by fitting a mechanistic, cellular dynamics-based tumor control probability that assumes a constant local supply of oxygen and glucose. In addition to standard radiobiological effects such as repair of sub-lethal damage and the impact of hypoxia, we also accounted for proliferation as well as radiosensitivity variability within the cell cycle. We applied the model to 36 published and two unpublished early-stage patient cohorts, totaling 2,701 patients.&lt;/p&gt;&lt;p&gt;&lt;b&gt;Results:&lt;/b&gt; Precise likelihood best-fit values were derived for the radiobiological parameters: α [0.305 Gy&lt;sup&gt;−1&lt;/sup&gt;; 95% confidence interval (CI), 0.120–0.365], the α/β ratio (2.80 Gy; 95% CI, 0.40–4.40), and the oxygen enhancement ratio (OER) value for intermediately hypoxic cells receiving glucose but not oxygen (1.70; 95% CI, 1.55–2.25). All fractionation groups are well fitted by a single dose–response curve with a high &lt;i&gt;χ&lt;sup&gt;2&lt;/sup&gt; P&lt;/i&gt; value, indicating consistency with the fitted model. The analysis was further validated with an additional 23 patient cohorts (&lt;i&gt;n&lt;/i&gt; = 1,628). The model indicates that hypofractionation regimens overcome hypoxia (and cell-cycle radiosensitivity variations) by the sheer impact of high doses per fraction, whereas lower dose-per-fraction regimens allow for reoxygenation and corresponding sensitization, but lose effectiveness for prolonged treatments due to proliferation.&lt;/p&gt;&lt;p&gt;&lt;b&gt;Conclusions:&lt;/b&gt; This proposed mechanistic tumor-response model can accurately predict overtreatment or undertreatment for various treatment regimens. &lt;i&gt;Clin Cancer Res; 23(18); 5469–79. ©2017 AACR&lt;/i&gt;.&lt;/p&gt;&lt;/div&gt;

&lt;div&gt;Abstract&lt;p&gt;Tissue survival responses to ionizing radiation are nonlinear with dose, rather yielding tissue-specific descending curves that impede straightforward analysis of biologic effects. Apoptotic cell death often occurs at low doses, while at clinically relevant intermediate doses, double-strand break misrepair yields mitotic death that determines outcome. As researchers frequently use a single low dose for experimentation, such strategies may inaccurately depict inherent tissue responses. Cutting edge radiobiology has adopted full dose survival profiling and devised mathematical algorithms to fit curves to observed data to generate highly reproducible numerical data that accurately define clinically relevant inherent radiosensitivities. Here, we established a protocol for irradiating organoids that delivers radiation profiles simulating the organ of origin. This technique yielded highly similar dose–survival curves of small and large intestinal crypts &lt;i&gt;in vivo&lt;/i&gt; and their cognate organoids analyzed by the single-hit multi-target (SHMT) algorithm, outcomes reflecting the inherent radiation profile of their respective Lgr5&lt;sup&gt;+&lt;/sup&gt; stem cell populations. As this technological advance is quantitative, it will be useful for accurate evaluation of intestinal (patho)physiology and drug screening.&lt;/p&gt;Significance:&lt;p&gt;These findings establish standards for irradiating organoids that deliver radiation profiles that phenocopy the organ of origin.&lt;/p&gt;&lt;p&gt;&lt;i&gt;See related commentary by Muschel et al., p. 927&lt;/i&gt;&lt;/p&gt;&lt;/div&gt;

&lt;p&gt;Table S1. Excluded cohorts in the analysis and the reason for exclusion; Table S2. The datasets used for this analysis in three different groups based on Mehta, et al.'s data (2012) and the estimated EQD22.8 (Gy) of the best-fit by the model simulation; Table S3. The additional datasets used for external validation of the model analysis; Figure S1. Deviations of estimated EQD210,model from BED; Figure S2. Dose-response curves estimated from maximum likelihood method based on the EQD210,model values; Figure S4. Dose-response curves obtained by applying other models to the dataset of current study: (A) Shuryak et al.'s model (69); and (B) Tai et al.'s model (70).&lt;/p&gt;

Supplementary Table 1 from A MicroRNA Expression Signature for Cervical Cancer Prognosis

Abstract Oncogenic driver mutations in different pediatric sarcoma subtypes have been identified but may not be druggable. In general, identifying novel therapeutic targets and biomarkers for response remains a major challenge. We hypothesize that considering the structure of the interaction network in which the genes operate as a system is crucial for understanding a gene's role. We propose to use the protein interaction network geometry to characterize the shape of network architecture and identify key aspects of direct and indirect cooperation pertaining to the cancer network and prognosis using geometrical methods. We model gene networks as weighted graphs where edges indicate protein-level interactions and edge weights estimate the strength of the interaction. The Human Protein Reference Database was used to define the gene network topology. RNA-Seq data from pediatric sarcoma tissues extracted from patients treated at MSK (n=12 Ewing sarcoma; n=29 osteosarcoma; n=20 desmoplastic small round cell tumor) was employed to prescribe correlation-based weights to create pediatric sarcoma subtype-specific weighted graphs. The geometry of the weighted gene networks was computed via a discrete notion of Ricci curvature. Intuitively, the curvature provides a measure of feedback (triangles) in the network. Positive curvature reflects robust communication and ease of information transfer, while negative curvature reflects bridge-like architecture or bottlenecks of information flow. We utilized a dynamic (multi-scale) notion of curvature to quantify the functional associations between genes, computed as a function of scale between diffusion processes initially localized on each node (i.e., gene). The curvature becomes more positive on edges between communal genes and more negative on bridge-like edges between communities, until reaching the critical scale. Curvature therefore, as we demonstrate, partitions the cancer networks into functionally associated communities. Community detection by removing bridge-edges, determined as edges with negative curvature at the critical scale, revealed sarcoma subtype-specific preferential gene associations. In particular, we agnostically found the EWSR1-FLI1 association in a cluster that was unique to the Ewing sarcoma network. Interestingly, we found ETV6 in the same community as the characteristic Ewing sarcoma EWSR1-FLI1 feature, suggesting a novel implication of ETV6 in Ewing sarcoma. These results suggest that persisting communities found by leveraging the cancer network geometry may identify potential mechanisms of drug resistance and actionable therapeutic targets. Citation Format: Rena Elkin, Jung Hun Oh, Filemon Dela Cruz, Larry Norton, Joseph O. Deasy, Andrew L. Kung, Allen R. Tannenbaum. Geometry of gene expression network reveals potential novel indicator in Ewing sarcoma [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 1 (Regular and Invited Abstracts); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(7_Suppl):Abstract nr 6541.

Abstract Tumors and healthy tissues exhibit gene expression variability in many of the same pathways. We propose that these commonalities result from a performance trade-off between universal cellular phenotypes; each phenotype is defined by an orthogonal gene expression profile identified using unsupervised dimensionality reduction. Using two publicly available RNA-seq datasets (n=54 normal tissues; n=1504 cancer cell lines), we show that both healthy and cancerous cells occupy a trade-off between five canonical phenotypes: OxPhos, Warburg, Fibroblastic, Immune, and Growth. Each cell is assigned a phenotype score that, as we demonstrate, predicts differential drug sensitivities and mutational signatures, even among cancers of the same tissue type. Since these phenotype scores are defined using only healthy tissues, they can be generally applied to other diseases as well. Citation Format: Corey Weistuch, Kevin Murgas, Ken Dill, Larry Norton, Joseph Deasy, Allen Tannenbaum. A universal atlas of cellular and oncogenic phenotypes [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 1 (Regular and Invited Abstracts); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(7_Suppl):Abstract nr 6591.

Abstract Cancer immune phenotypes present a wide range of heterogeneity across cases, with individual tumors displaying unique patterns of infiltrating immune cell types. Deconvolutional methods allow for scoring of various immune cell types in bulk tumor RNA as a quantification of immune phenotype. Understanding how immune phenotype relates to clinical outcome remains limited. Here, we demonstrate an approach applying topological data analysis to investigate differences of immune phenotype in a pan-cancer cohort (TCGA; n=11,373 tumors). We first define an Immune Activation Score based on relative abundance of activator and suppressor immune cell types and find this score depends on cancer type and distinguishes overall survival outcomes. We then implement a robust Mapper-based algorithm to delineate clusters of immune phenotypes of tumor samples across pan-cancer and within cancer types. Our method identifies immune-activated and immune-suppressed phenotypes with distinct survival outcomes and molecular features. Citation Format: Kevin A. Murgas, Jung H. Oh, Joseph O. Deasy, Allen R. Tannenbaum. Topological data analysis reveals pan-cancer immune phenotypes with immune-related survival differences. [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 1 (Regular and Invited Abstracts); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(7_Suppl):Abstract nr 4657.

Abstract Multiple myeloma (MM) prognosis incorporates a variety of metrics including treatment received, clinical factors, and genomic characteristics, with several specific genomic features predicting for shorter progression free survival (PFS). No study to date has integrated genomic data from a systems view, incorporating interactions between biomarkers in a network. We use a geometric network analysis that integrates complex interactions to characterize patterns of biological behavior not captured by individual genomic events. The methodology is mathematically well-defined and has no fitting parameters. We hypothesized that such a systems mathematical approach applied to gene interaction networks may delineate biologically relevant MM subtypes and potential new therapeutic targets. We overlaid RNA-Seq and copy number alteration data from the MMRF CoMMpass study (IA19) on a gene interactome derived from the Human Protein Reference Database using a novel graph metric of network robustness — Ollivier-Ricci curvature (ORC). Results were clustered, with the optimal number determined via silhouette score. Survival analysis for PFS was performed employing Kaplan-Meier and log-rank tests. A differential gene expression analysis between high and low risk groups was conducted. Differences in scalar ORC between the low-risk and high-risk groups were examined and contextualized using a pathway analysis. Pathway analysis was performed using the Broad Institute’s Gene Set Enrichment Analysis tool and the pathways used are from the hallmark gene set from the Human MSigDB collection. The dataset included 659 patients and the incorporated protein-protein interactions resulted in a network with 8,468 nodes and 33,695 edges. The ORC analysis discovered 6 clusters, with specific genomic features being associated with clusters predicting for long [hyperdiploidy, t(11:14)], and short [t(4;14), MAF/MAFB translocations] PFS. A differential gene expression analysis comparing the high risk and low risk groups identified 118 key genes. These genes were associated with various pathways both known and unknown to be associated with multiple myeloma, including mitotic spindle, DNA repair, inflammatory response, and the P53 pathways. Further scalar curvature analysis showed differences in the apoptosis, TGF beta signaling, and other signaling pathways. In summary, we applied the geometric network analysis tool ORC to multi-omics data in MM represented as biological networks to identify individuals at high risk of short PFS and relevant biological correlates. Decreased robustness of signaling near immune-related genes was associated with shorter survival, highlighting the plausible utility of using these methods to uncover new biological insights. Citation Format: Anish K. Simhal, Kylee H. Maclachlan, Rena Elkin, Jiening Zhu, Saad Z. Usmani, Jonathan J. Keats, Larry Norton, Joseph O. Deasy, Jung Hun Oh, Allen Tannenbaum. Protein network analysis uncovers a poor-survival subtype in multiple myeloma [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 1 (Regular and Invited Abstracts); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(7_Suppl):Abstract nr 2061.

Abstract The modern development of sequencing technologies provides a comprehensive molecular portrait of human cancers. There is a strong need to develop methods to not only improve patient prognosis predictions but also to understand the driving factors for treatment. However, the high-dimension, low-sample size nature of the genomic data poses challenges for typical machine learning algorithms. The systematic understanding of genes with respect to a network (protein-protein interaction (PPI) network) is a way to handle the limit and the nonparametric analysis of geometric properties such as Ollivier-Ricci curvature and associated invariant measure developed by our group have proven to be successful for the prediction of survival in multiple cancers. In this work, we propose a novel supervised deep learning approach combining the aforementioned geometric methods, which benefit from the flexibility provided by deep learning techniques while still preserving much of the interpretability of the geometric analysis. We take advantage of a state-of-the-art graph neural network approach. Sparse connections between layers were inspired by the known biology of the PPI network from the Human Protein Reference Database (HPRD) and pathway information from the Kyoto Encyclopedia of Genes and Genomes (KEGG) database, supplemented with geometric network features which are fed into the network in corresponding layers. The prediction is based on a local-global principle, where highly predictive features are selected from early layers of the network and fed directly to the final layer to produce a multivariable Cox regression. We applied our method to RNA-Seq gene expression data from the CoMMpass study of multiple myeloma (MM). More specifically, 657 patients in the data set were randomly divided into training, validation and set-aside testing sets by a ratio of 6:2:2. We obtained an average C-index 0.66 of the prediction in the testing set from a 10-fold data split. Dichotomizing the testing set by its mean value to define high-risk vs. low-risk yielded a significant p-value of the log-rank test in the set-aside data (p-value =3e-4). We observed that geometric protein network information not only improved the outcome prediction (vs. 6% worse without geometric feature inputs), but was also more robust to fold splitting. From our model, we identified WEE1, CENPE and CENPF as top genes driving survival differences (higher expression of WEE1 increased risk and lower negative curvature between CENPE and CENPF increased risk). WEE1 is a cell cycle-related gene that regulates DNA repair and CENPE and CENPF are components of a fibrous layer of mitotic kinetochores, which have been indicated in the literature to be related to the prognosis as well as possible targets for treatment. While it is therefore logical that these genes would be implicated in the natural history of MM, they were identified entirely on the basis of network analysis. Citation Format: Jiening Zhu, Jung Hun Oh, Anish K. Simhal, Rena Elkin, Larry Norton, Joseph O. Deasy, Allen R. Tannenbaum. Deep neural networks using protein-protein network information predict multiple myeloma survival. [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 1 (Regular and Invited Abstracts); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(7_Suppl):Abstract nr 5367.

Drug-induced liver injury (DILI) is an adverse hepatic drug reaction that can potentially lead to life-threatening liver failure. Previously published work in the scientific literature on DILI has provided valuable insights for the understanding of hepatotoxicity as well as drug development. However, the manual search of scientific literature in PubMed is laborious and time-consuming. Natural language processing (NLP) techniques along with artificial intelligence/machine learning approaches may allow for automatic processing in identifying DILI-related literature, but useful methods are yet to be demonstrated. To address this issue, we have developed an integrated NLP/machine learning classification model to identify DILI-related literature using only paper titles and abstracts. For prediction modeling, we used 14,203 publications provided by the Critical Assessment of Massive Data Analysis (CAMDA) challenge, employing word vectorization techniques in NLP in conjunction with machine learning methods. Classification modeling was performed using 2/3 of the data for training and the remainder for test in internal validation. The best performance was achieved using a linear support vector machine (SVM) model on the combined vectors derived from term frequency-inverse document frequency (TF-IDF) and Word2Vec, resulting in an accuracy of 95.0% and an F1-score of 95.0%. The final SVM model constructed from all 14,203 publications was tested on independent datasets, resulting in accuracies of 92.5%, 96.3%, and 98.3%, and F1-scores of 93.5%, 86.1%, and 75.6% for three test sets (T1-T3). Furthermore, the SVM model was tested on four external validation sets (V1-V4), resulting in accuracies of 92.0%, 96.2%, 98.3%, and 93.1%, and F1-scores of 92.4%, 82.9%, 75.0%, and 93.3%.

Patient outcomes with definitive CRT for LA-NSCLC remain poor, with no imaging biomarkers to predict benefit. Hence, we developed a serial image AI model using paired planning CT (pCT) and first week cone-beam CT (CBCT) to predict PFS and measured AI model fairness defined as the bias in the classification with respect to gender as a protected attribute.Sixty-four consecutive patients with LA-NSCLC treated with concurrent CRT to 60 Gy in 30 fractions and durvalumab consolidation were analyzed. Three prediction models were created. A previously developed AI image foundation model [1] was pre-trained with unlabeled 6,402 3D CT scans sourced from institutional and the Cancer Imaging Archive and modified to predict PFS as a binarized outcome (high PFS > 6 months and low PFS < 6 months) using pCT scans. Serial image AI model was created by adding the first week CBCT scan. The third model measured tumor growth rate (TGR) as relative change in tumor and nodal volume from pCT to CBCT derived using a different published AI model [2]. Association with PFS using univariable and multivariable Cox regression after adjusting for age, gender, planning tumor volume, and smoking status were measured using TGR and the two AI model predictions using a cutoff of > 50% probability for low PFS. AI model fairness metrics area under receiver operating curve (AUROC), precision, sensitivity, and specificity were computed.TGR was not associated with PFS on univariate (Hazard ratio [HR] of 1.515, 95% confidence interval [CI] of 0.32 to 7.26, p = 0.60) or multivariate analysis (HR: 1.58, 95% CI: 0.32 to 7.80, p = 0.58) and resulted in a Harrell's C-index of 54.7%. The serial image AI model prediction was associated with PFS in both univariable (HR: 2.12, 95% CI: 1.02 to 4.40, p = 0.045) and multivariable analysis (HR 2.39, 95% CI of 1.09 to 5.25, p = 0.029), and a C-index of 62.5%. The pCT AI model was associated with PFS in univariate (HR 2.06, 95% CI of 1.06 to 4.01, p = 0.034) but not in multivariable analysis (HR 1.89, 95% CI of 0.93 to 3.87, p = 0.08), and a C-index of 59.9%. The serial image AI model reduced the parity in classification compared to pCT AI model indicating higher fairness (Table I).The multi-image AI model predicted PFS with slightly higher accuracy and resulted in higher fairness than the pCT AI model. These results underscore the potential for incorporating multi-imaging biomarkers to predict treatment response.

Purpose/Objective(s)Prostate cancer localization during radiotherapy is difficult and introduces positional variability during treatment. Here we evaluate the impact on treatment toxicities of an in-house system that uses MV and kV imaging guidance (MKIG) to track implanted fiducials during stereotactic body radiotherapy (SBRT) for localized prostate cancer.Materials/MethodsA 3D MKIG platform that tracks prostate implanted fiducials in real-time was built and clinically translated to replace a prior commercial approach called intrafraction motion review (IMR), which tracks thefiducials in 2D kV view only. MKIG has been shown to correct both superior-inferior and anterior-posterior (AP) motions that are harmful to critical organs and is superior to IMR for intrafractional motion management, which is less sensitive to AP motions. From 2017 to 2019, 150 patients with localized prostate cancer were treated with SBRT to 40 Gy in 5 fractions. During the delivery of volumetric modulated arc therapy, orthogonal MV-kV pairs were simultaneously acquired at every 20° gantry rotation and rigidly registered to the reference image templates created from the planning CT. Calculated 3D translations of implanted fiducials were used to localize the prostate and alert the therapist to interrupt and reposition the prostate when exceeding a 1.5-2 mm threshold. A comparison cohort of 121 prostate patients was treated from 2015 to 2016 with the same prescription dose and treatment technique but instead managed by IMR, where the therapist interrupted treatment based on a 2mm expansion of the fiducial contours superimposed on the kV images. Statistics of intrafractional interruptions, patient shifts, and overall delivery time were collected to evaluate the efficacy of the clinical workflow. The incidence of late grade ≥2 toxicities was analyzed to assess clinical complications. The median follow-up time was 5.5 years (range of 3.6 to 8.0 years).ResultsThe MKIG cohort had more interruptions per fraction (1.09 vs. 0.28) and longer average delivery time per fraction (579±205s vs. 357±117s) than IMR. 75% of shifts resulting from MKIG were less than 3mm, compared to 51% in IMR, indicating that MKIG tended to detect and correct smaller deviations (p<0.001). The baseline International Prostate Symptom Score was 7.9 in the MKIG cohort vs. 8.4 in IMR (p = 0.41). The incidence of late grade ≥2 urinary toxicity was lower in the MKIG than IMR cohort: 10.7% vs. 19.8% (p = 0.05). One grade ≥2 rectal toxicity was observed in the IMR cohort but none in MKIG.ConclusionWehave demonstrated that MKIG is a clinically practical and effective method for monitoring and correcting prostate positional deviations during SBRT of prostate cancer. MKIG is better suited than 2D IMR to localize the prostate and trigger patient repositioning during treatment. A statistically and clinically significant reduction in urinary toxicity was observed. The potential expansion of MKIG to other clinical sites and translation to other centers should be considered. Prostate cancer localization during radiotherapy is difficult and introduces positional variability during treatment. Here we evaluate the impact on treatment toxicities of an in-house system that uses MV and kV imaging guidance (MKIG) to track implanted fiducials during stereotactic body radiotherapy (SBRT) for localized prostate cancer. A 3D MKIG platform that tracks prostate implanted fiducials in real-time was built and clinically translated to replace a prior commercial approach called intrafraction motion review (IMR), which tracks thefiducials in 2D kV view only. MKIG has been shown to correct both superior-inferior and anterior-posterior (AP) motions that are harmful to critical organs and is superior to IMR for intrafractional motion management, which is less sensitive to AP motions. From 2017 to 2019, 150 patients with localized prostate cancer were treated with SBRT to 40 Gy in 5 fractions. During the delivery of volumetric modulated arc therapy, orthogonal MV-kV pairs were simultaneously acquired at every 20° gantry rotation and rigidly registered to the reference image templates created from the planning CT. Calculated 3D translations of implanted fiducials were used to localize the prostate and alert the therapist to interrupt and reposition the prostate when exceeding a 1.5-2 mm threshold. A comparison cohort of 121 prostate patients was treated from 2015 to 2016 with the same prescription dose and treatment technique but instead managed by IMR, where the therapist interrupted treatment based on a 2mm expansion of the fiducial contours superimposed on the kV images. Statistics of intrafractional interruptions, patient shifts, and overall delivery time were collected to evaluate the efficacy of the clinical workflow. The incidence of late grade ≥2 toxicities was analyzed to assess clinical complications. The median follow-up time was 5.5 years (range of 3.6 to 8.0 years). The MKIG cohort had more interruptions per fraction (1.09 vs. 0.28) and longer average delivery time per fraction (579±205s vs. 357±117s) than IMR. 75% of shifts resulting from MKIG were less than 3mm, compared to 51% in IMR, indicating that MKIG tended to detect and correct smaller deviations (p<0.001). The baseline International Prostate Symptom Score was 7.9 in the MKIG cohort vs. 8.4 in IMR (p = 0.41). The incidence of late grade ≥2 urinary toxicity was lower in the MKIG than IMR cohort: 10.7% vs. 19.8% (p = 0.05). One grade ≥2 rectal toxicity was observed in the IMR cohort but none in MKIG. Wehave demonstrated that MKIG is a clinically practical and effective method for monitoring and correcting prostate positional deviations during SBRT of prostate cancer. MKIG is better suited than 2D IMR to localize the prostate and trigger patient repositioning during treatment. A statistically and clinically significant reduction in urinary toxicity was observed. The potential expansion of MKIG to other clinical sites and translation to other centers should be considered.

The Wasserstein distance from optimal mass transport (OMT) is a powerful mathematical tool with numerous applications that provides a natural measure of the distance between two probability distributions. Several methods to incorporate OMT into widely used probabilistic models, such as Gaussian or Gaussian mixture, have been developed to enhance the capability of modeling complex multimodal densities of real datasets. However, very few studies have explored the OMT problems in a reproducing kernel Hilbert space (RKHS), wherein the kernel trick is utilized to avoid the need to explicitly map input data into a high-dimensional feature space. In the current study, we propose a Wasserstein-type metric to compute the distance between two Gaussian mixtures in a RKHS via the kernel trick, i.e., kernel Gaussian mixture models.

Lung adenocarcinoma (ADC) is the most common non-small cell lung cancer. Surgical resection is the primary treatment for early-stage lung ADC while lung-sparing surgery is an alternative for non-aggressive cases. Identifying histopathologic subtypes before surgery helps determine the optimal surgical approach. Predominantly solid or micropapillary (MIP) subtypes are aggressive and associated with a higher likelihood of recurrence and metastasis and lower survival rates. This study aims to non-invasively identify these aggressive subtypes using preoperative 18F-FDG PET/CT and diagnostic CT radiomics analysis. We retrospectively studied 119 patients with stage I lung ADC and tumors ≤ 2 cm, where 23 had aggressive subtypes (18 solid and 5 MIPs). Out of 214 radiomic features from the PET/CT and CT scans and 14 clinical parameters, 78 significant features (3 CT and 75 PET features) were identified through univariate analysis and hierarchical clustering with minimized feature collinearity. A combination of Support Vector Machine classifier and Least Absolute Shrinkage and Selection Operator built predictive models. Ten iterations of 10-fold cross-validation (10×10-fold CV) evaluated the model. A pair of texture feature (PET GLCM Correlation) and shape feature (CT Sphericity) emerged as the best predictor. The radiomics model significantly outperformed the conventional predictor SUVmax (accuracy: 83.5% vs. 74.7%, p=9e-9) and identified aggressive subtypes by evaluating FDG uptake asymmetry in the tumor. It also demonstrated a high negative predictive value of 95.6% compared to SUVmax (88.2%, p=2e-10). The proposed radiomics approach could reduce unnecessary extensive surgeries for non-aggressive subtype patients, improving surgical decision-making for early-stage lung ADC patients.

Abstract Melanoma response to immune-modulating therapy remains incompletely characterized at the molecular level. In this study, we assess melanoma immunotherapy response using a multi-scale network approach to identify gene modules with coordinated gene expression in response to treatment. Using gene expression data of melanoma before and after treatment with nivolumab, we modeled gene expression changes in a correlation network and measured a key network geometric property, dynamic Ollivier-Ricci curvature, to distinguish critical edges within the network and reveal multi-scale treatment-response gene communities. Analysis identified six distinct gene modules corresponding to sets of genes interacting in response to immunotherapy. One module alone, overlapping with the nuclear factor kappa-B pathway (NFKB), was associated with improved patient survival and a positive clinical response to immunotherapy. This analysis demonstrates the usefulness of dynamic Ollivier-Ricci curvature as a general method for identifying information-sharing gene modules in cancer.

The plasma cell cancer multiple myeloma (MM) varies significantly in genomic characteristics, response to therapy, and long-term prognosis. Recent studies have shown the complexity of biological systems associated with poor outcomes in MM. One hallmark of cancer pathogenesis, applicable to MM, is the dysregulation of apoptotic cell death. Understanding the relationship between known drivers of MM and critical pathways in the apoptosis gene network is key towards identifying novel therapeutic targets. Recently, there has been a surge of interest in developing network-based approaches to study gene interactions and use machine learning tools to further elucidate useful biological factors. It is well-known that genes operate as network systems; however, interpretable computational methods are still limited. In this work, we combine a set of network analysis tools to provide novel biological insights into how the apoptosis network is dysregulated in MM. We identify a set of genes associated with poor outcomes in MM using both univariate and network-based approaches. Genomic and clinical data, including RNA-sequencing (RNA-seq) and copy number alteration (CNA), for 659 subjects were obtained from the Multiple Myeloma Research Foundation's CoMMpass (IA19) database. The apoptosis gene set is defined using the apoptosis hallmark pathway provided by the gene set enrichment analysis tool. The interactions between the given list of genes are provided via MetaCore. Univariate gene modeling is done using a Cox proportional hazards model. To determine the relationship between genes, we used Ollivier-Ricci curvature (ORC), a measure of network robustness. The ORC value between two genes incorporates information about both the two genes in question and their local neighborhoods, a concept critical for understanding the role a neighborhood of genes has on a given connection. Once edge information is computed, we used a uniform manifold approximation and projection (UMAP) to reduce the complexity of the dataset. The resulting projection is clustered using k-means clustering, and survival analysis of the given clusters is done using the Kaplan-Meier (KM) method. To understand which features are most informative, we used a random forest to predict the cluster labels using the RNA-seq data. After five-fold cross-validation, feature importance was computed using random forest's permutation importance methodology. All statistical significance testing was corrected for multiple comparisons via the Benjamini-Hochberg false discovery rate method with an alpha set to 0.05. Using RNA-seq data of genes associated with apoptosis, we identified nineteen apoptosis genes associated with poor prognosis in MM. Some of these genes, such as ANXA1 and WEE1 are known to have an effect in MM, while others, such as AVPR1A have yet to be studied extensively for their role in MM. When examining CNA data associated with apoptosis, a similar pattern emerged - nineteen genes were associated with poor prognosis. However, between the two key gene sets found by RNA-seq and CNA, only FAS appeared in both lists. FAS, Fas Cell Surface Death Receptor, is critical for triggering apoptosis. Half of the 37 genes identified using RNA-seq and CNA data are directly connected to the TP53 gene. While TP53 mutations are not a marker of the presence of MM itself, TP53 mutations have been shown to be associated with poorer outcomes. After reducing the complexity of the Ollivier-Ricci curvature analysis of the RNA-seq apoptosis network using UMAPs, the data revealed five clusters with KM curves for progression-free survival were significant (p&amp;lt;0.001). To identify the genes most impactful for separating these clusters, we trained a random forest model on the RNA-seq - note, it was the edge robustness values that were clustered - and evaluated its performance for predicted class membership. The model had an overall accuracy of approximately 78% with F1 scores for each class ranging from 61% to 87%. The top three features associated with the model include CTNNB1, RELA, and LEF1. Of note, CTNNB1 and RELA are directly interacted with the TP53 gene. Because these three genes discriminate between all five clusters, their expression levels may play a mediating effect on MM outcomes. Further investigation with external datasets is needed to validate these results.

In August 2022, the Cancer Informatics for Cancer Centers brought together cancer informatics leaders for its biannual symposium, Precision Medicine Applications in Radiation Oncology, co-chaired by Quynh-Thu Le, MD (Stanford University), and Walter J. Curran, MD (GenesisCare). Over the course of 3 days, presenters discussed a range of topics relevant to radiation oncology and the cancer informatics community more broadly, including biomarker development, decision support algorithms, novel imaging tools, theranostics, and artificial intelligence (AI) for the radiotherapy workflow. Since the symposium, there has been an impressive shift in the promise and potential for integration of AI in clinical care, accelerated in large part by major advances in generative AI. AI is now poised more than ever to revolutionize cancer care. Radiation oncology is a field that uses and generates a large amount of digital data and is therefore likely to be one of the first fields to be transformed by AI. As experts in the collection, management, and analysis of these data, the informatics community will take a leading role in ensuring that radiation oncology is prepared to take full advantage of these technological advances. In this report, we provide highlights from the symposium, which took place in Santa Barbara, California, from August 29 to 31, 2022. We discuss lessons learned from the symposium for data acquisition, management, representation, and sharing, and put these themes into context to prepare radiation oncology for the successful and safe integration of AI and informatics technologies.

&lt;div&gt;Abstract&lt;p&gt;Invasive cervical cancer is a leading cause of cancer death in women worldwide, resulting in about 300,000 deaths each year. The clinical outcomes of cervical cancer vary significantly and are difficult to predict. Thus, a method to reliably predict disease outcome would be important for individualized therapy by identifying patients with high risk of treatment failures before therapy. In this study, we have identified a microRNA (miRNA)-based signature for the prediction of cervical cancer survival. miRNAs are a newly identified family of small noncoding RNAs that are extensively involved in human cancers. Using an established PCR-based miRNA assay to analyze 102 cervical cancer samples, we identified miR-200a and miR-9 as two miRNAs that could predict patient survival. A logistic regression model was developed based on these two miRNAs and the prognostic value of the model was subsequently validated with independent cervical cancers. Furthermore, functional studies were done to characterize the effect of miRNAs in cervical cancer cells. Our results suggest that both miR-200a and miR-9 could play important regulatory roles in cervical cancer control. In particular, miR-200a is likely to affect the metastatic potential of cervical cancer cells by coordinate suppression of multiple genes controlling cell motility. Cancer Res; 70(4); 1441–8&lt;/p&gt;&lt;/div&gt;

&lt;div&gt;Abstract&lt;p&gt;Invasive cervical cancer is a leading cause of cancer death in women worldwide, resulting in about 300,000 deaths each year. The clinical outcomes of cervical cancer vary significantly and are difficult to predict. Thus, a method to reliably predict disease outcome would be important for individualized therapy by identifying patients with high risk of treatment failures before therapy. In this study, we have identified a microRNA (miRNA)-based signature for the prediction of cervical cancer survival. miRNAs are a newly identified family of small noncoding RNAs that are extensively involved in human cancers. Using an established PCR-based miRNA assay to analyze 102 cervical cancer samples, we identified miR-200a and miR-9 as two miRNAs that could predict patient survival. A logistic regression model was developed based on these two miRNAs and the prognostic value of the model was subsequently validated with independent cervical cancers. Furthermore, functional studies were done to characterize the effect of miRNAs in cervical cancer cells. Our results suggest that both miR-200a and miR-9 could play important regulatory roles in cervical cancer control. In particular, miR-200a is likely to affect the metastatic potential of cervical cancer cells by coordinate suppression of multiple genes controlling cell motility. Cancer Res; 70(4); 1441–8&lt;/p&gt;&lt;/div&gt;

To independently test a previously published model to predict radiation pneumonitis and to further test the correlation reported therein of heart dose/volume irradiation factors with radiation pneumonitis (RP). All patients with heterogeneity-corrected dose distributions in recoverable treatment planning archives, treated for stage II or III NSCLC with definitive fractionated radiation therapy between November 2004 and January 2010 were included (98 datasets). All normal structures were re-contoured in accordance with published guidelines. Lung and heart dosimetric parameters were extracted using CERR (The Computational Environment for Radiation therapy Research). Available clinical data included patient age, gender, race, smoking, weight loss, T stage, N stage, fraction size, chemotherapy, and tumor location parameters. Forty-six patients (47%) developed Grade 2 or worse RP: 36/Grade 2, 9/Grade 3, and 1/Grade 4. Statistically significant variables with the highest univariate Spearman rank correlations included: heart V40 (0.22, p < 0.02), heart D20 (0.17, p < 0.05), heart D30 (0.17, p < 0.05), pre-treatment chemotherapy (0.16, p < 0.06), lung V60 (0.16, p < 0.07), lung D15 (0.13, p < 0.1), and lung D20 (0.14, p < 0.09). The three-variable RP model proposed in Huang et al (which uses heart D10, lung D35, and lung maximum dose), was able to predict RP risk in the current dataset with a spearman rank correlation coefficient of 0.22 (p < 0.02), which is slightly lower than the coefficient reported for the older dataset (0.268) but still significant. Irradiation of a sizeable fraction of the heart to more than about 40 Gy confers a risk of RP, in the new dataset as well as in the original dataset analyzed in Huang et al. Once again, a range of heart dose-volume parameters were more predictive of RP than all lung parameters. However, these results are likely to be dependent on the types of treatments delivered at this institution. Also, the results do not preclude the possibility that a more predictive model can be found.

Recent data has shown that for oropharyngeal squamous cell carcinoma (OPSCC), p16 positivity (p16+), a surrogate for HPV, portends a better prognosis. At Washington University, the most frequent approach for OPSCC is minimally invasive transoral laser microsurgery (TLM) with adjuvant radiotherapy (RT) or chemoradiotherapy (CRT). This retrospective review compares outcomes for OPSCC p16+ patients treated with adjuvant RT or CRT. A total of 376 consecutive patients with OPSCC were available for analysis from the years 1997-2009, with 220 receiving adjuvant IMRT. of the 139 adjuvant IMRT patients whose p16 and chemotherapy status were known, 100 were p16+, of which 52 were treated with RT and 48 with CRT. Platinum-based chemotherapy was delivered to 46 patients and cetuximab to 2 patients ineligible for chemotherapy. The tumor bed and previously involved lymph node levels were treated with 60 to 66 Gy at 2.0 Gy/fraction. The uninvolved N0 side of the neck received 52 to 56 Gy at 1.6 to 1.73 Gy/fraction. A propensity score analysis was performed using the following variables: age, gender, site, stage, grade, Adult Comorbidity Evaluation-27 (ACE-27) index, extracapsular extension, T stage, and N stage. The median follow-up was 47 months. The median age at diagnosis was 55 and 54 years in the RT and CRT groups, respectively. The majority of patients had stage III and IV OPSCC, with 48 patients in the RT and CRT groups each. Extracapsular extension was observed in 86.3% of patients, and positive margins in 14% of patients. For all patients, the locoregional recurrence free survival rate was 96%. Overall survival (OS) was 89% and 81% at 2 and 5 years, respectively. There were 3 locoregional failures in the RT group and 1 in the CRT group. All 4 failures had evidence of extracapsular extension, and 2 had close margins while 2 had negative margins. OS for the RT group was 92% (95% CI = 80 - 97%) and 87% (95% CI = 73 - 94%) at 2 and 5 years, respectively. OS for the CRT group was 85% (95% CI = 74 - 94%) and 74% (95% CI = 56 - 86%) at 2 and 5 years, respectively. No statistically significant differences between these groups were seen in either OS (p = 0.58) or DFS (p = 0.39; stratified log-rank analysis) based on the propensity score analysis. Patients with p16+, HPV associated OPSCC undergoing surgery and adjuvant IMRT have low rates of locoregional failures despite high rates of extracapsular extension. The role of chemotherapy was unclear in this retrospective analysis since it was not powered to detect significant, clinically important differences. Further studies are necessary to elucidate the role of adjuvant chemotherapy in p16+ OPSCC.

According to classical radiobiology, the cellular effects of ionizing radiation are attributed to radiation-induced DNA damage. However, radiation can also produce effects in non-targeted cells by a presumed immunological mechanism. Although the mechanism is unknown, the involvement of oxidative and inflammatory responses is considered likely pivotal. We hypothesized that partial tumor irradiation may allow for immune activation and ideal sparing of nearby critical normal structures limiting toxicity. We examined the responses of tumors that were partially exposed to single high dose radiation therapy (SDRT) to either the entire tumor or half the tumor in immunocompetent Balb/c mice and nude mice implanted with 67NR murine breast tumors. Immunohistochemistry was performed as well as antibody-mediated depletion of CD8-positive T cells. In Balb/c mice partial irradiation of tumors with SDRT led to a tumor response similar to the ones fully exposed to irradiation. Using exposure to 20Gy led to endothelial apoptosis and decreased microvascular density, both in irradiated and non-irradiated parts of the tumor. While 10Gy SDRT did not have a marked toxic effect on endothelial cells or profound DNA damage measured by ɣH2AX compared to 20 Gy, it induced an increased expression of ICAM and E-selectin, and infiltration of cytotoxic T-cells to both irradiated and non-irradiated parts of the tumor. Irradiation of 50% of the tumor area with 10Gy in these mice led to the same tumor response as irradiation of 100% of the tumor. However, mice, treated to 100% of the tumor ultimately succumbed to GI toxicity. Response in 50% irradiated tumors was abrogated by treatment with anti-CD8 antibody, indicating the involvement of T cells in this tumor response. For immunogenic tumors suboptimal tumor coverage due to dose constraints may still be potentially curative in a murine model via activation of an anti-tumor immune response. Future experiments will probe the underlying mechanism as well as effect of varying dose characteristics.

Purpose: We have developed a novel small animal radiation therapy device (microRT), which integrates multi‐modality imaging, radiation treatment planning, and conformal radiation therapy. In this study, we evaluated the accuracy of the treatment planning and positioning systems of the microRT device. Method and Materials: The microRT system utilizes a clinical 192 Ir HDR source collimated via machined tungsten inserts to deliver photon beams at a source to target distances of 1–8cm at four angles (0, 90, 180, and 270). Beams were modeled using Monte Carlo and a parameterized analytic dose engine was created. Radiochromic film (5mm steps) in a solid water phantom was used to evaluate actual delivered doses in multiple planes. Treatment plans using these beams were created by a custom treatment planning system (microRTP) based on imported fiducial‐registered imaging (CT, MR, PET) of animals immobilized in the treatment position. A three‐axis computer‐controlled stage supports and positions animals in the beams according to the microRTP plan. Validation of the positioning system was performed using a phantom and images of phantom and collimator via a kV C‐arm. Results: The analytic dose model agreed with the Monte‐Carlo predicted dose within 5% and 10% outside and inside the 1 mm deep build‐up regions, respectively. Film dosimetry agreed with the analytic model within 10% and also demonstrated an effective field diameter of 8mm at 17mm from the source. The 192 Ir line source geometry caused a radial anisotropy of up to 12% at 17 mm depth from the source. The positioning accuracy of the animal support hardware was sub‐millimeter. Conclusions: The microRT system provides conformal radiation therapy based on pre‐treatment imaging and planning for small animal models of cancer and tissue injury. This work supported in part by NIH R21 CA108677 and by a grant from Varian, Inc.

Clinical trials and oncology data management have undergone considerable change in the past decade. Imaging has become a key tool for clinical trials management and a biomarker for clinical trial validation as imaging technologies improve and become more precise. Images have become extremely helpful in determining staging/eligibility, treatment response, and outcome determination including disease recurrence and progression. In modern protocols, images are often reviewed in real time to validate these points in order to improve compliance to study requirements and create uniform patient populations for clinical trials analysis. Data acquisition and management systems are currently in use to acquire and display images in electronic digital formats for view by both on site and off site radiology reviewers. As clinical trials become more global in focus, the ability for databases to accommodate diverse imaging acquisition strategies will become increasingly important for information review.

Purpose: Radiotherapy treatment of lung cancer patients is complicated by changes in tumor position due to breathing and changes in size due to regression. Accurate quantification of these changes during the course of treatment would likely improve tumor response and reduce toxicity risks. We are investigating robust methods for tracking and estimating tumor volume changes between treatment fractions. Method and Materials: We have developed a registration—assisted segmentation approach based on the level‐set deformable model, in which pre‐treatment contours are propagated and adapted to fractions times at selected respiratory phases. At any time‐point during treatment, a reference respiratory phase is selected and corresponding 3D‐CT volumes are reconstructed from 4D‐CT acquisition data. Images are then globally aligned using an efficient registration algorithm. Pre‐treatment planning contours are copied to selected time‐points. In our tumor regression analysis, the GTV contour was used to initialize the level set algorithm in‐place and the PTV contour was used to narrowband the region, thus improving the algorithm's convergence. The feasibility of the proposed approach was investigated on a set of patients with repeated 4D scans acquired at three time‐intervals. Results: Our preliminary analysis indicates that the proposed approach can properly capture the boundary of the shrinking tumor region or split regions due to its topological adaptation ability. On an initial cohort of four NSCLC patients, the estimated tumor volume reductions ranged between 3–46% with a median of 8% by mid‐treatment and between 26–51% with a median of 34% by the end of treatment. Conclusion: We have demonstrated a new approach for tracking tumor regression during the course of radiotherapy treatment of NSCLC patients based on a novel level‐set segmentation algorithm. This approach provides us with a semi‐automated tool for quantifying tumor shrinkage and allows accurate estimates of ‘true’ accumulated dose to the tumor. Supported by CA85181 grant.

We hypothesize that there may be a differential effect on the risk of radiation pneumonitis (RP) from irradiating bronchial vs. non-bronchial regions within normal lung. We studied this potential effect within a multivariate framework, for patients treated with radiation therapy for lung cancer. All evaluable, archived 3-D treatment plans for patients with registered outcomes treated for NSCLC between 1991 and 2001 were eligible. RP which resulted in steroid use or more intensive intervention was classified as an event (WUSTL Grade 2 or higher; RTOG Grade 3 or higher). Doses were retrospectively corrected for heterogeneity effects via a Monte Carlo-based method. Plans were processed and reviewed using CERR (computational environment for radiotherapy research). Heart volumes were re-contoured within CERR by a single physician (n = 209, with 48 RP events). The structure of bronchi in lung for each available patient was auto-contoured using active contour methods implemented in the ITK-SNAP software package. Heart and normal lung (lung minus gross tumor volume) dose-volume parameters were extracted for further modeling using CERR. Evaluated factors included clinical (age, gender, race, performance status, weight loss, smoking, histology); dosimetric parameters for heart (D5-D100, V10-V80, mean dose, maximum dose, and minimum dose); treatment factors (chemotherapy, treatment time, fraction size); and location parameters (heart-center-of-dose, sup-inf within the heart; GTV location within the lung). Dose-volume parameters were also separately extracted for bronchial vs. non-bronchial normal lung regions. The best model order was determined using leave-one-out cross validation. The best multivariate model was obtained by step-wise variable selection and logistic regression. Leave-one-out cross validation analysis supported a model order of four variables (with rank correlation coefficients on the out-of-sample points of about 0.31). The most commonly selected variables were (in order of decreasing frequency): heart D10 (minimum dose to the hottest 10%), heart D30, bronchi V80, and non-bronchi V20. Although heart dose-volume parameters were more statistically significant, the RP risk model improved by including bronchial and non-bronchial dose-volume parameters separately. A high-dose bronchial irradiation term frequently was selected, indicating that RP-risk may be increased when bronchi are irradiated to relatively high doses. These correlations should be tested further against new datasets.

To detect and quantify, within a multivariate framework, the effect of heart irradiation on the risk of radiation pneumonitis (RP) for radiotherapy patients treated for lung cancer. All evaluable, archived 3D treatment plans for patients with registered outcomes treated for NSCLC between 1991 and 2001 were eligible. RP which resulted in steroid use or more intensive intervention was classified as an event (WUSTL Grade 2 or higher; RTOG Grade 3 or higher). Doses were retrospectively corrected for heterogeneity effects via a Monte Carlo-based method. Plans were processed and reviewed using CERR (computational environment for radiotherapy research). Heart volumes were re-contoured within CERR by a single physician (n = 209, with 48 RP events). Heart and normal lung (lung minus gross tumor volume) dose-volume parameters were extracted for further modeling using CERR. Evaluated factors included clinical (age, gender, race, performance status, weight loss, smoking, histology); dosimetric parameters for heart (D5-D100, V10-V80, mean dose, maximum dose, and minimum dose); treatment factors (chemotherapy, treatment time, fraction size); location parameters (heart center-of-dose, sup-inf within the heart); as well as previously identified significant lung parameters: mean lung dose and sup-inf tumor position (GTV center-of-mass) within the lung (Bradley JD et al., IJROBP 2007;69(4):985-992). A best multivariate model was obtained by step-wise variable selection and logistic regression. The best model order was determined using leave-one-out cross validation. Statistically significant variables with the highest univariate Spearman rank correlations included: maximum heart dose (0.227, p < 0.0006), heart V70 (0.239, p < 0.0003), heart D5 (minimum dose to the hottest 5% of the heart) (0.256, p < 0.0002), heart D10 (0.24, p < 0.0003), heart gEUD (0.249 for a = 10, p < 0.0001), and GTV_COMSI (0.219, p < 0.0008). Leave-one-out cross-validation analysis supported a model order of four variables (with rank correlation coefficients on the out-of-sample points of about 0.3). The most commonly selected variables were (in order of decreasing frequency): heart D10, heart D30, mean lung dose, and GTV_COMSI within the lung. Heart dose-volume parameters were more statistically significant than any previously derived lung parameters within a multivariate modeling framework to predict RP. These correlations should be tested further against new datasets.

Abstract We introduce a classification of breast tumors into 7 classes which are more clearly defined by interpretable mRNA signatures along the PAM50 gene set than the 5 traditional PAM50 intrinsic subtypes. Each intrinsic subtype is partially concordant with one of our classes, and the 2 additional classes correspond to division of the classes concordant with the Luminal B and the Normal intrinsic subtypes along expression of the Her2 gene group. Our Normal class shows similarity with the myoepithelial mammary cell phenotype, including TP63 expression (specificity: 80.8% and sensitivity: 82.8%), and exhibits the best overall survival (89.6% at 5 years). Though Luminal A tumors are traditionally considered the least aggressive, our analysis shows that only the Luminal A tumors which are now classified as myoepithelial have this phenotype, while tumors in our luminal class (concordant with Luminal A) may be more aggressive than previously thought. We also find that patients with Basal tumors surviving to 48 months exhibit favorable survival rates when certain markers for B-lymphocytes are present and poor survival rates when they are absent, which is consistent with recent findings.

State-of-the-Art in Radiomics in Radiology and Radiation Oncology Radiomics is the science of converting medical images into mineable data, data that are descriptive of “phenotypes,” which may provide diagnostic, prognostic, or therapeutic information. Genomics is the science of sequencing and analyzing the function and structure of genomes; the complete set of DNA in a single cell of an organism. In turn, imaging genomics (or radiogenomics) is concerned with the correlation between image-based features, as determined by radiomics, and gene expression, as determined by genomics. Imaging genomics arose from decades of work in at least three key areas: 1) sequencing of the human genome, 2) quantitative imaging, computer-aided diagnosis, therapy prognostics, and assessment of therapy response in preclinical and clinical research and practice; and 3) data science, including the burgeoning areas of genomics, preclinical and population-based disease modeling, individualized medicine, and big data. Imaging genomics may answer important questions in medicine by correlating validated, quantitative image phenotypes (through validated imaging biomarkers) with clinical data, histopathologic data, molecular classifications, genomic assays, and treatment outcomes. This approach could address some of the greatest health burdens, including cancer, cardiac disease, and arthritis. Participants will discuss the state-of-the-art of radiomics across multiple disease sites and modalities. Aspects of the presentations will include how to improve the quality of image-based phenotypes of normal and diseased tissue, how to better determine the relationships between these phenotypes and the underlying biology associated with the images, and how to create predictive models using the image-based phenotypes. Topics will also include: 1) creating, archiving, curating and sharing ultra-large datasets (“big data”); 2) standardizing image acquisition and processing methods; 3) standardizing and validating phenotype extraction methods and classifier designs; and 4) using high-throughput, robust, and validated phenotyping systems. H. Aerts, NIH; NCIH. Li, This research was funded in part by the University of Chicago Dean Bridge Fund,and by NCI U24-CA143848-05, P50- CA58223 Breast SPORE program. Hui Li received royalties from Hologic.W. Lu, This work was supported in part by the National Cancer Institute Grants R01CA172638.

Purpose: The aim of this study was to investigate the usefulness of the QUANTEC guideline to prevent xerostomia after intensity‐modulated radiotherapy (IMRT) for head and neck cancer (HNC) with respect to follow‐up time. In addition, we explored alternative guidelines to further reduce xerostomia. Methods: The QUANTEC guideline suggests a mean dose to the contralateral (Dmeancontra) parotid&lt;20 Gy, or Dmeancontra and Dmean to the ipsilateral parotid (Dmeanipsi)&lt;25 Gy. Stimulated whole mouth saliva flow measurements (WMSFM) were conducted at a median of 11 (3–24) months for 63 patients treated with IMRT for HNC to a median dose of 70 Gy in 2006–2015. Severe hyposalivation/xerostomia was defined as WMSFM ≤25% post‐ relative to pre‐RT. Patients were stratified into a &lt;6m (xerostomia: 27% (n=15)), and a 6–24m (xerostomia: 19% (n=10)) follow‐up group. Dose‐response modeling was performed using logistic regression including Dmeancontra, or Dmeancontra and Dmeanipsi. The area under the receiver‐operating characteristics curve (AUC) was used to assess discriminative ability, and the agreement between the estimated and observed rate of xerostomia was given by Spearman's rank correlation coefficient (Rs), and related p‐values. Results: Of the non‐xerostomia patients, 65% (&lt;6m) and 56% (6–24m) fulfilled Dmeancontra&lt;20 Gy. The estimated and observed rate of xerostomia agreed at &lt;6m (AUC=0.78; Rs=0.46; p=0.001), and was 28% at Dmeancontra=20 Gy. A smaller number of non‐xerostomia patients fulfilled the two‐gland guideline (33% (&lt;6m) and 26% (6–24m)), but the AUC was higher than using Dmeancontra only (&lt;6m: AUC=0.90; Rs=0.63; p&lt;0.0001; 6–24m: AUC=0.84; Rs=0.25; p=0.08), and the following amendment of the two‐gland guideline was suggested: (0.17*Dmeancontra+0.11*Dmeanipsi‐8.13)&lt;‐1.60 (&lt;6m), and (0.05*Dmeancontra+0.02*Dmeanipsi‐3.10)&lt;‐1.60 (6–24m). Conclusion: The QUANTEC guideline is effective to prevent xerostomia &lt;6m post‐RT, but its usefulness is reduced at later follow‐up times. The suggested amendment to the two‐gland QUANTEC guideline should be further investigated in an independent cohort of HNC patients treated with IMRT.