An 18F-FDG PET/CT and Mean Lung Dose Model to Predict Early Radiation Pneumonitis in Stage III Non-Small Cell Lung Cancer Patients Treated with Chemoradiation and Immunotherapy

Radiation pneumonitis (RP) that develops early (i.e., within 3 mo) (RPEarly) after completion of concurrent chemoradiation (cCRT) leads to treatment discontinuation and poorer survival for patients with stage III non-small cell lung cancer. Since no RPEarly risk model exists, we explored whether published RP models and pretreatment 18F-FDG PET/CT-derived features predict RPEarly Methods: One hundred sixty patients with stage III non-small cell lung cancer treated with cCRT and consolidative immunotherapy were analyzed for RPEarly Three published RP models that included the mean lung dose (MLD) and patient characteristics were examined. Pretreatment 18F-FDG PET/CT normal-lung SUV featured included the following: 10th percentile of SUV (SUVP10), 90th percentile of SUV (SUVP90), SUVmax, SUVmean, minimum SUV, and SD. Associations between models/features and RPEarly were assessed using area under the receiver-operating characteristic curve (AUC), P values, and the Hosmer-Lemeshow test (pHL). The cohort was randomly split, with similar RPEarly rates, into a 70%/30% derivation/internal validation subset. Results: Twenty (13%) patients developed RPEarly Predictors for RPEarly were MLD alone (AUC, 0.72; P = 0.02; pHL, 0.87), SUVP10, SUVP90, and SUVmean (AUC, 0.70-0.74; P = 0.003-0.006; pHL, 0.67-0.70). The combined MLD and SUVP90 model generalized in the validation subset and was deemed the final RPEarly model (RPEarly risk = 1/[1+e(- x )]; x = -6.08 + [0.17 × MLD] + [1.63 × SUVP90]). The final model refitted in the 160 patients indicated improvement over the published MLD-alone model (AUC, 0.77 vs. 0.72; P = 0.0001 vs. 0.02; pHL, 0.65 vs. 0.87). Conclusion: Patients at risk for RPEarly can be detected with high certainty by combining the normal lung's MLD and pretreatment 18F-FDG PET/CT SUVP90 This refined model can be used to identify patients at an elevated risk for premature immunotherapy discontinuation due to RPEarly and could allow for interventions to improve treatment outcomes.

Multi-scale geometric network analysis identifies melanoma immunotherapy response gene modules

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.

Functional transcriptional signatures for tumor-type-agnostic phenotype prediction

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.

Wasserstein HOG: Local Directionality Extraction via Optimal Transport

Directionally sensitive radiomic features including the histogram of oriented gradient (HOG) have been shown to provide objective and quantitative measures for predicting disease outcomes in multiple cancers. However, radiomic features are sensitive to imaging variabilities including acquisition differences, imaging artifacts and noise, making them impractical for using in the clinic to inform patient care. We treat the problem of extracting robust local directionality features by mapping via optimal transport a given local image patch to an iso-intense patch of its mean. We decompose the transport map into sub-work costs each transporting in different directions. To test our approach, we evaluated the ability of the proposed approach to quantify tumor heterogeneity from magnetic resonance imaging (MRI) scans of brain glioblastoma multiforme, computed tomography (CT) scans of head and neck squamous cell carcinoma as well as longitudinal CT scans in lung cancer patients treated with immunotherapy. By considering the entropy difference of the extracted local directionality within tumor regions, we found that patients with higher entropy in their images, had significantly worse overall survival for all three datasets, which indicates that tumors that have images exhibiting flows in many directions may be more malignant. This may seem to reflect high tumor histologic grade or disorganization. Furthermore, by comparing the changes in entropy longitudinally using two imaging time points, we found patients with reduction in entropy from baseline CT are associated with longer overall survival (hazard ratio = 1.95, 95% confidence interval of 1.4-2.8, p = 1.65e-5). The proposed method provides a robust, training free approach to quantify the local directionality contained in images.

Dynamic network curvature analysis of gene expression reveals novel potential therapeutic targets in sarcoma

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.

"Primer shot" fractionation with an early treatment break is theoretically superior to consecutive weekday fractionation schemes for early-stage non-small cell lung cancer

PURPOSE

Radiotherapy is traditionally given in equally spaced weekday fractions. We hypothesize that heterogeneous interfraction intervals can increase radiosensitivity via reoxygenation. Through modeling, we investigate whether this minimizes local failures and toxicity for early-stage non-small cell lung cancer (NSCLC).

METHODS

Previously, a tumor dose-response model based on resource competition and cell-cycle-dependent radiosensitivity accurately predicted local failure rates for early-stage NSCLC cohorts. Here, the model mathematically determined non-uniform inter-fraction intervals minimizing local failures at similar normal tissue toxicity risk, i.e., iso-BED3 (iso-NTCP) for fractionation schemes 18Gyx3, 12Gyx4, 10Gyx5, 7.5Gyx8, 5Gyx12, 4Gyx15. Next, we used these optimized schedules to reduce toxicity risk (BED3) while maintaining stable local failures (TCP).

RESULTS

Optimal schedules consistently favored a "primer shot" fraction followed by a 2-week break, allowing tumor reoxygenation. Increasing or decreasing the assumed baseline hypoxia extended or shortened this optimal break by up to one week. Fraction sizes of 7.5 Gy and up required a single primer shot, while smaller fractions needed one or two extra fractions for full reoxygenation. The optimized schedules, versus consecutive weekday fractionation, predicted absolute LF reductions of 4.6%-7.4%, except for the already optimal LF rate seen for 18Gyx3. Primer shot schedules could also reduce BED3 at iso-TCP with the biggest improvements for the shortest schedules (94.6Gy reduction for 18Gyx3).

CONCLUSION

A validated simulation model clearly supports non-standard "primer shot" fractionation, reducing the impact of hypoxia-induced radioresistance. A limitation of this study is that primer-shot fractionation is outside prior clinical experience and therefore will require clinical studies for definitive testing.

Exploring published and novel pre-treatment CT and PET radiomics to stratify risk of progression among early-stage non-small cell lung cancer patients treated with stereotactic radiation

PURPOSE

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.

MATERIALS/METHODS

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.

RESULTS

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).

CONCLUSIONS

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.

Artificial intelligence-based automated segmentation and radiotherapy dose mapping for thoracic normal tissues

Background and purpose

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.

Materials and methods

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).

Results

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).

Conclusions

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.

Auto-segmentation of neck nodal metastases using self-distilled masked image transformer on longitudinal MR images

Objectives

Auto-segmentation promises greater speed and lower inter-reader variability than manual segmentations in radiation oncology clinical practice. This study aims to implement and evaluate the accuracy of the auto-segmentation algorithm, "Masked Image modeling using the vision Transformers (SMIT)," for neck nodal metastases on longitudinal T2-weighted (T2w) MR images in oropharyngeal squamous cell carcinoma (OPSCC) patients.

Methods

This prospective clinical trial study included 123 human papillomaviruses (HPV-positive [+]) related OSPCC patients who received concurrent chemoradiotherapy. T2w MR images were acquired on 3 T at pre-treatment (Tx), week 0, and intra-Tx weeks (1-3). Manual delineations of metastatic neck nodes from 123 OPSCC patients were used for the SMIT auto-segmentation, and total tumor volumes were calculated. Standard statistical analyses compared contour volumes from SMIT vs manual segmentation (Wilcoxon signed-rank test [WSRT]), and Spearman's rank correlation coefficients (ρ) were computed. Segmentation accuracy was evaluated on the test data set using the dice similarity coefficient (DSC) metric value. P-values <0.05 were considered significant.

Results

No significant difference in manual and SMIT delineated tumor volume at pre-Tx (8.68 ± 7.15 vs 8.38 ± 7.01 cm3, P = 0.26 [WSRT]), and the Bland-Altman method established the limits of agreement as -1.71 to 2.31 cm3, with a mean difference of 0.30 cm3. SMIT model and manually delineated tumor volume estimates were highly correlated (ρ = 0.84-0.96, P < 0.001). The mean DSC metric values were 0.86, 0.85, 0.77, and 0.79 at the pre-Tx and intra-Tx weeks (1-3), respectively.

Conclusions

The SMIT algorithm provides sufficient segmentation accuracy for oncological applications in HPV+ OPSCC.

Advances in knowledge

First evaluation of auto-segmentation with SMIT using longitudinal T2w MRI in HPV+ OPSCC.

Cardiac radiation dose is associated with inferior survival but not cardiac events in patients with locally advanced non-small cell lung cancer in the era of immune checkpoint inhibitor consolidation

PURPOSE

We assessed the association of cardiac radiation dose with cardiac events and survival post-chemoradiation therapy (CRT) in patients with locally advanced non-small cell lung cancer (LA-NSCLC) after adoption of modern radiation therapy (RT) techniques, stricter cardiac dose constraints, and immune checkpoint inhibitor (ICI) consolidation.

METHODS AND MATERIALS

This single-institution, multi-site retrospective study included 335 patients with LA-NSCLC treated with definitive, concurrent CRT between October 2017 and December 2021. All patients were evaluated for ICI consolidation. Planning dose constraints included heart mean dose < 20 Gy (<10 Gy if feasible) and heart volume receiving ≥ 50 Gy (V50Gy) < 25 %. Twenty-one dosimetric parameters for three different cardiac structures (heart, left anterior descending coronary artery [LAD], and left ventricle) were extracted. Primary endpoint was any major adverse cardiac event (MACE) post-CRT, defined as acute coronary syndrome, heart failure, coronary revascularization, or cardiac-related death. Secondary endpoints were: grade ≥ 3 cardiac events (per CTCAE v5.0), overall survival (OS), lung cancer-specific mortality (LCSM), and other-cause mortality (OCM).

RESULTS

Median age was 68 years, 139 (41 %) had baseline coronary heart disease, and 225 (67 %) received ICI consolidation. Proton therapy was used in 117 (35 %) and intensity-modulated RT in 199 (59 %). Median LAD V15Gy was 1.4 % (IQR 0-22) and median heart mean dose was 8.7 Gy (IQR 4.6-14.4). Median follow-up was 3.3 years. Two-year cumulative incidence of MACE was 9.5 % for all patients and 14.3 % for those with baseline coronary heart disease. Two-year cumulative incidence of grade ≥ 3 cardiac events was 20.4 %. No cardiac dosimetric parameter was associated with an increased risk of MACE or grade ≥ 3 cardiac events. On multivariable analysis, cardiac dose (LAD V15Gy and heart mean dose) was associated with worse OS, driven by an association with LCSM but not OCM.

CONCLUSIONS

With modern RT techniques, stricter cardiac dose constraints, and ICI consolidation, cardiac dose was associated with LCSM but not OCM or cardiac events in patients with LA-NSCLC.

Stereotactic Body Radiation Therapy for Stage IIA to IIIA Inoperable Non-Small Cell Lung Cancer: A Phase 1 Dose-Escalation Trial

PURPOSE

Larger tumors are underrepresented in most prospective trials on stereotactic body radiation therapy (SBRT) for inoperable non-small cell lung cancer (NSCLC). We performed this phase 1 trial to specifically study the maximum tolerated dose (MTD) of SBRT for NSCLC >3 cm.

METHODS AND MATERIALS

A 3 + 3 dose-escalation design (cohort A) with an expansion cohort at the MTD (cohort B) was used. Patients with inoperable NSCLC >3 cm (T2-4) were eligible. Select ipsilateral hilar and single-station mediastinal nodes were permitted. The initial SBRT dose was 40 Gy in 5 fractions, with planned escalation to 50 and 60 Gy in 5 fractions. Adjuvant chemotherapy was mandatory for cohort A and optional for cohort B, but no patients in cohort B received chemotherapy. The primary endpoint was SBRT-related acute grade (G) 4+ or persistent G3 toxicities (Common Terminology Criteria for Adverse Events version 4.03). Secondary endpoints included local failure (LF), distant metastases, disease progression, and overall survival.

RESULTS

The median age was 80 years; tumor size was >3 cm and ≤5 cm in 20 (59%) and >5 cm in 14 patients (41%). In cohort A (n = 9), 3 patients treated to 50 Gy experienced G3 radiation pneumonitis (RP), thus defining the MTD. In the larger dose-expansion cohort B (n = 25), no radiation therapy-related G4+ toxicities and no G3 RP occurred; only 2 patients experienced G2 RP. The 2-year cumulative incidence of LF was 20.2%, distant failure was 34.7%, and disease progression was 54.4%. Two-year overall survival was 53%. A biologically effective dose (BED) <100 Gy was associated with higher LF (P = .006); advanced stage and higher neutrophil/lymphocyte ratio were associated with greater disease progression (both P = .004).

CONCLUSIONS

Fifty Gy in 5 fractions is the MTD for SBRT to tumors >3 cm. A higher BED is associated with fewer LFs even in larger tumors. Cohort B appears to have had less toxicity, possibly due to the omission of chemotherapy.

Interpretable Machine Learning for Choosing Radiation Dose-volume Constraints on Cardio-pulmonary Substructures Associated with Overall Survival in NRG Oncology RTOG 0617

PURPOSE

Our objective was to use interpretable machine learning for choosing dose-volume constraints on cardiopulmonary substructures (CPSs) associated with overall survival (OS) in radiation therapy for locally advanced non-small cell lung cancer.

METHODS AND MATERIALS

A total of 428 patients with non-small cell lung cancer were randomly divided into training/validation/test subsets (n = 230/149/49) in Radiation Therapy Oncology Group 0617. Manual or automated contouring was performed to segment CPSs, including heart, atria, ventricles, aorta, left/right ventricle/atrium (LV+RV+LA+RA), inferior/superior vena cava, pulmonary artery, and pericardium. Peri (pericardium-heart), rest (heart-[LV+RV+LA+RA]), clinical target volume (CTV), and lungs-CTV contours were also obtained. Dose-volume histogram features were extracted, including minimum/mean dose to the hottest x% volume (Dx%[Gy]/MOHx%[Gy]), minimum/mean/maximum dose, percent volume receiving at least xGy (VxGy[%]), and overlapping volume of each CPS with planning target volume (PTV_Voverlap[%]). Clinical parameters were collected from the National Clinical Trials Network/Community oncology research program data archive. Feature selection was performed using a series of multiblock sparse partial least squares regression, stability selection supervised principal component analysis, and Boruta. Explainable boosting machine (EBM) was trained using a conditional survival distribution-based approach for imputing censored data, treating survival analysis as a regression problem. Harrell's C-index was used to evaluate OS discrimination performance of EBM, Cox proportional hazards (CPH), random survival forest, extreme gradient boosting survival embeddings, and CPH deep neural network (DeepSurv) models in the test set. Dose-volume constraints were selected using the binary change point detection algorithm in Shapley additive explanations-based partial dependence functions.

RESULTS

Selected features included LA_V60Gy(%), pericardium_D30%(Gy), lungs-CTV_PTV_Voverlap(%), RA_V55Gy(%), and received_cons_chemo. All models ranked LA_V60Gy(%) as the most important feature. EBM achieved the best performance for predicting OS, followed by extreme gradient boosting survival embeddings, random survival forest, DeepSurv, and CPH (C-index = 0.653, 0.646, 0.642, 0.638, and 0.632). EBM global explanations suggested that LA_V60Gy(%) < 25.6, lungs-CTV_PTV_Voverlap(%) < 1.1, pericardium_D30%(Gy) < 18.9, RA_V55Gy(%) < 19.5, and received_cons_chemo = 'Yes' for improved OS.

CONCLUSIONS

EBM can be used to discriminate OS while also guiding dose-volume constraint selection for optimal management of cardiac toxicity in lung cancer radiation therapy.

ROE (Radiotherapy Outcomes Estimator): An open-source tool for optimizing radiotherapy prescriptions

BACKGROUND AND OBJECTIVES

Radiotherapy prescriptions currently derive from population-wide guidelines established through large clinical trials. We provide an open-source software tool for patient-specific prescription determination using personalized dose-response curves.

METHODS

We developed ROE, a plugin to the Computational Environment for Radiotherapy Research to visualize predicted tumor control and normal tissue complication simultaneously, as a function of prescription dose. ROE can be used natively with MATLAB and is additionally made accessible in GNU Octave and Python, eliminating the need for commercial licenses. It provides a curated library of published and validated predictive models and incorporates clinical restrictions on normal tissue outcomes. ROE additionally provides batch-mode tools to evaluate and select among different fractionation schemes and analyze radiotherapy outcomes across patient cohorts.

CONCLUSION

ROE is an open-source, GPL-copyrighted tool for interactive exploration of the dose-response relationship to aid in radiotherapy planning. We demonstrate its potential clinical relevance in (1) improving patient awareness by quantifying the risks and benefits of a given treatment protocol (2) assessing the potential for dose escalation across patient cohorts and (3) estimating accrual rates of new protocols.

Gene interaction network analysis in multiple myeloma detects complex immune dysregulation associated with shorter survival

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.

Multi-Scale Geometric Network Analysis Identifies Melanoma Immunotherapy Response Gene Modules

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.

Activation of STING in Response to Partial-Tumor Radiation Exposure

PURPOSE

To determine the mechanisms involved in partial volume radiation therapy (RT)-induced tumor response.

METHODS AND MATERIALS

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.

RESULTS

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.

CONCLUSIONS

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.

Preoperative 18F-FDG PET/CT and CT radiomics for identifying aggressive histopathological subtypes in early stage lung adenocarcinoma

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 in the tumor and tumor shape. 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.

Simulating the Potential of Model-Based Individualized Prescriptions for Ultracentral Lung Tumors

Purpose

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.

Methods and Materials

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.

Results

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).

Conclusions

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.

Association of Cardiac Dose with Cardiac Events and Survival for Locally Advanced Non-Small Cell Lung Cancer (LA-NSCLC) Treated with Concurrent Chemoradiotherapy (cCRT) in the Era of Immune Checkpoint Inhibitor (ICI) Consolidation

PURPOSE/OBJECTIVE(S)

To assess the association of cardiac dose with post-cCRT cardiac events and survival among patients (pts) with LA-NSCLC after adoption of ICI consolidation, modern radiotherapy (RT) techniques, and data-driven cardiac constraints.

MATERIALS/METHODS

This single-institution, multi-site retrospective study included 335 pts with LA-NSCLC treated with definitive cCRT (60-70 Gy) from October 2017 to December 2021. Pts were evaluated for ICI consolidation. Cardiac dose constraints included heart volume receiving ≥50 Gy (V50) <25% and mean heart dose (MHD) <20 Gy. Heart, left anterior descending artery (LAD), and left ventricle were autocontoured, manually reviewed, and edited. 21 dosimetric parameters (mean dose, max dose, and min dose to the hottest x% volume [Dx%(Gy); x from 5-95 in 5% intervals]) for each were extracted, as well as LAD V15. Baseline cardiovascular disease (bCVD) was defined as heart failure (HF), coronary artery disease, peripheral vascular disease, or cerebrovascular disease. Primary endpoint was post-cCRT major adverse cardiac events (MACE), defined as acute coronary syndrome, HF hospitalization/urgent visit, coronary revascularization, or cardiac death. Secondary endpoints were grade ≥3 cardiac events (CTCAE v5.0), overall survival (OS), cancer specific mortality (CSM), and other cause mortality (OCM). Competing risk regression was used for MACE and grade ≥3 cardiac events, and Cox regression for OS, CSM, and OCM.

RESULTS

Median age was 68 years, 139 (41%) had bCVD, and 225 (67%) received consolidation ICI. Proton therapy was used in 117 (35%), intensity-modulated RT in 199 (59%), and 3D conformal RT in 19 (6%). Median MHD was 8.7 Gy (IQR 4.6-14.4) and median LAD V15 1.4% (IQR 0-22). Median follow-up was 39.5 months. 35 MACE events occurred; 1- and 2-year cumulative incidence (CI) were 4.2% and 9.5%. No cardiac dosimetric parameter associated with MACE after adjusting for bCVD and age (e.g., MHD sHR 0.98/Gy, 95% CI 0.93-1.03, p = 0.43) or within the following 3 subgroups: no bCVD, photon therapy, and ICI consolidation. 87 grade ≥3 cardiac events occurred; 1- and 2- year CI were 12.6% and 20.4%. Heart dose was not associated with grade ≥3 cardiac events after adjusting for bCVD, ECOG, and BMI (e.g., MHD sHR 1.00/Gy, 95% CI 0.97-1.03, p = 0.85) or within the 3 aforesaid subgroups. 183 OS events occurred, including 125 CSM and 58 OCM events. Multiple cardiac dosimetric parameters associated with worse OS on multivariable analysis (e.g., LAD V15 HR 1.01/%, 95% CI 1.00-1.02, p = 0.003), driven by associations with CSM (LAD V15 HR 1.02/%, p<0.001) but not OCM (LAD V15 HR 1.00/%, p = 0.73). Median OS was worse for LAD V15 ≥10% (22.2 vs 35.1 months, p = 0.004).

CONCLUSION

Among pts with LA-NSCLC treated with cCRT after adoption of ICI consolidation, modern RT techniques, and cardiac constraints, post-cCRT cardiac events were common but showed no association with cardiac dose. Cardiac dose associated with OS, driven by an association with CSM and not OCM, which may not reflect cardiac toxicity.

Multitask AI Models for the Joint Prediction of Overall Survival, Progression-Free Survival, and Death without Progression as a Composite Endpoint for LA-NSCLC Patients Treated with Chemoradiotherapy

PURPOSE/OBJECTIVE(S)

Prior methods model the risk of endpoints separately. Herein, we construct a composite AI model that considers multiple endpoints jointly, including overall survival (OS), progression-free survival (PFS), and death without progression (DWP). Our hypothesis is that the composite model potentially improves predictive performance for patients with locally advanced non-small cell lung cancer (LANSCLC) treated with chemoradiotherapy (CRT).

MATERIALS/METHODS

A total of 335 LANSCLC patients treated with definitive CRT, including all evaluable patients accrued from Oct 2017 to Dec 2021, were randomly split into training/test subsets (n = 234/101). Cardio-pulmonary substructures (CPSs) were autocontoured, manually reviewed, and edited if necessary. A total of 1093 non-independent dosimetric parameters were extracted, including GTVp, GTVn, GTV, PTV, esophagus, lungs minus IGTV, left/right lung, 15 CPSs, and the overlapping volume of each OAR with PTV and the distance from each OAR to GTVp/GTVn. Other clinical parameters included age, consolidation immunotherapy (CI), ECOG score, Charlson comorbidity index, coronary heart disease, histology, PD-L1 expression, and clinical stage (AJCC 8). Within training, censored time-to-event data were imputed based on conditional event distributions derived from Kaplan-Meier estimators for casting survival analysis as a regression problem and training neural additive model (NAM) regressors. Features were selected by LASSO regression for a single endpoint (OS, PFS, DWP) and multi-task (MT) LASSO regression for four separate composite endpoints (OS-PFS, OS-DWP, PFS-DWP, OS-PFS-DWP). The performance of MT NAMs in the test set that jointly predicted the composite endpoints was evaluated using the C-index and compared to that of a single task (ST) NAM that predicted each endpoint separately.

RESULTS

The best testing performance in predicting OS and DWP was attained by the MT NAM that jointly predicted all endpoints (c-index = 0.65, 95% CI 0.58-0.71 for OS; c-index = 0.78, 95% CI 0.69-0.87 for DWP). The best model to predict PFS was also MT between PFS and DWP (c-index = 0.59, 95% CI 0.52-0.65). The c-indices of all ST NAMs were less than 0.56. The best MT NAMs significantly outperformed ST NAMs in predicting OS (p = 0.001) and DWP (p = 0.01) except for PFS (p = 0.32). The best MT NAM in predicting OS and DWP included ECOG score, atria-PTV overlap volume, D75% [Gy] to the left atrium (LA), pulmonary arterial volume, histology (adenocarcinoma), D65% [Gy] to the descending aorta (DA), V10 Gy [%] of the LA and CI in order of overall importance. ECOG score consistently ranked as the most important feature for all four MT NAMs. An increase of ECOG score from 0 to 2 indicated a 6-month earlier risk of mortality and DWP. Atria-PTV overlap volume and D65% [Gy] to the DA were included in all four MT NAMs.

CONCLUSION

MT AI models improved outcome prediction in patients with LANSCLC treated with CRT by jointly learning commonalities between the primary and auxiliary endpoints.

Reduction of Late Urinary Toxicity from Prostate Cancer Radiotherapy via Intrafractional MV-kV Image Guidance

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/METHODS

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).

RESULTS

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.

CONCLUSION

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.

AI Serial Image Prediction of Progression-Free Survival (PFS) for Locally Advanced Non-Small Cell Lung Cancer (LA-NSCLC) Patients Treated with Chemoradiation (CRT) and Durvalumab Consolidation

PURPOSE/OBJECTIVE(S)

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.

MATERIALS/METHODS

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.

RESULTS

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).

CONCLUSION

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.

A Novel FDG PET and Mean Lung Dose Model to Identify Stage III NSCLC Patients at High Risk of Developing Early Radiation Pneumonitis

PURPOSE/OBJECTIVE(S)

Early radiation pneumonitis (RPEarly) is a primary reason for the premature discontinuation of durvalumab consolidation and can lead to poor survival in patients with stage III non-small cell lung cancer (NSCLC). Currently, there are no RP risk models specifically tested for RPEarly. Here, we tested the applicability of published RP models for predicting RPEarly and explored the value of integrating pretreatment FDG-PET parameters.

MATERIALS/METHODS

The cohort consisted of all 178 LA-NSCLC patients treated with concurrent chemoradiation (cCRT) and durvalumab between May 2017 and December 2021. RPEarly was defined as RP occurring within three months of cCRT completion; late RP (RPLate) was defined as any later occurring RP. The three published RP models analyzed included: 1) Mean lung dose (MLD), 2) MLD, age, pulmonary comorbidity, smoking status, and tumor location, and 3) MLD, age and pulmonary comorbidity. In addition, pretreatment FDG PET-CT scans were used to calculate SUV parameters from auto-segmented normal lung contours: 10th- and 90th percentile (SUVP10, SUVP90), maximum, mean (SUVmean), minimum, and standard deviation. The RP models were fit to RPEarly, RPLate, and RPEarly+Late in the 178 patients. To assess the association between FDG PET parameters and RP unbiasedly, the cohort was then randomly split, but enforcing similar RP rates, into a two-thirds derivation and a one-third validation subset. Model performance was assessed by AUC, p-values and the Hosmer-Lemeshow test (pHL; ideally ∼0.50).

RESULTS

The rates of RPEarly, RPLate, and RPEarly+Late were 12%, 11%, and 23%, respectively (corresponding to 21, 20, and 41 events). Only the MLD model significantly predicted RPEarly (AUC = 0.70; p = 0.04; pHL = 0.84); none of the three models predicted RPLate or RPEarly+Late. Among the FDG PET parameters, SUVP10, SUVP90 and SUVmean predicted RPEarly with similar performance (AUC = 0.69-0.73; p = 0.005-0.01; pHL = 0.68-0.72), and, therefore, bivariate models were built between MLD and each of SUVP10, SUVP90 and SUVmean. Only the MLD + SUVP90 model generalized in the validation subset (AUC = 0.63; p = 0.03; pHL = 0.89) and was thus deemed the final model for RPEarly. A final re-fitting of all model coefficients to the whole cohort indicated improvement over using the published MLD alone model (AUC = 0.75 vs. 0.70; p-value = 0.0006 vs. 0.04; pHL = 0.67 vs. 0.84). Risk of RPEarly is thus estimated as: RPEarly = 1/(1 = e-x); x = -5.79 + (1.57*MLD) + (0.14* SUVP90).

CONCLUSION

Patients at risk for RPEarly can be accurately identified prior to treatment by combining a re-fitted version of the published Mean Lung Dose model and pre-treatment FDG PET SUVP90 of the normal lung. This refined model can be used to identify patients with an exacerbated risk for premature durvalumab discontinuation due to RPEarly and could allow for interventions and/or the generation of "RPEarly sparing" treatment plans to improve overall treatment outcomes.

Mapping the Single-cell Differentiation Landscape of Osteosarcoma

The genetic and intratumoral heterogeneity observed in human osteosarcomas (OS) poses challenges for drug development and the study of cell fate, plasticity, and differentiation, processes linked to tumor grade, cell metastasis, and survival. To pinpoint errors in OS differentiation, we transcriptionally profiled 31,527 cells from a tissue-engineered model that directs MSCs toward adipogenic and osteoblastic fates. Incorporating pre-existing chondrocyte data, we applied trajectory analysis and non-negative matrix factorization (NMF) to generate the first human mesenchymal differentiation atlas. This 'roadmap' served as a reference to delineate the cellular composition of morphologically complex OS tumors and quantify each cell's lineage commitment. Projecting these signatures onto a bulk RNA-seq OS dataset unveiled a correlation between a stem-like transcriptomic phenotype and poorer survival outcomes. Our study takes the critical first step in accurately quantifying OS differentiation and lineage, a prerequisite to better understanding global differentiation bottlenecks that might someday be targeted therapeutically.

Geometric graph neural networks on multi-omics data to predict cancer survival outcomes

The advance of sequencing technologies has enabled a thorough molecular characterization of the genome in human cancers. To improve patient prognosis predictions and subsequent treatment strategies, it is imperative to develop advanced computational methods to analyze large-scale, high-dimensional genomic data. However, traditional machine learning methods face a challenge in handling the high-dimensional, low-sample size problem that is shown in most genomic data sets. To address this, our group has developed geometric network analysis techniques on multi-omics data in connection with prior biological knowledge derived from protein-protein interactions (PPIs) or pathways. Geometric features obtained from the genomic network, such as Ollivier-Ricci curvature and the invariant measure of the associated Markov chain, have been shown to be predictive of survival outcomes in various cancers. In this study, we propose a novel supervised deep learning method called geometric graph neural network (GGNN) that incorporates such geometric features into deep learning for enhanced predictive power and interpretability. More specifically, we utilize a state-of-the-art graph neural network with sparse connections between the hidden layers based on known biology of the PPI network and pathway information. Geometric features along with multi-omics data are then incorporated into the corresponding layers. The proposed approach utilizes a local-global principle in such a manner that highly predictive features are selected at the front layers and fed directly to the last layer for multivariable Cox proportional-hazards regression modeling. The method was applied to multi-omics data from the CoMMpass study of multiple myeloma and ten major cancers in The Cancer Genome Atlas (TCGA). In most experiments, our method showed superior predictive performance compared to other alternative methods.

Predicting the germline dependence of hematuria risk in prostate cancer radiotherapy patients

BACKGROUND AND PURPOSE

Late radiation-induced hematuria can develop in prostate cancer patients undergoing radiotherapy and can negatively impact the quality-of-life of survivors. If a genetic component of risk could be modeled, this could potentially be the basis for modifying treatment for high-risk patients. We therefore investigated whether a previously developed machine learning-based modeling method using genome-wide common single nucleotide polymorphisms (SNPs) can stratify patients in terms of the risk of radiation-induced hematuria.

MATERIALS AND METHODS

We applied a two-step machine learning algorithm that we previously developed for genome-wide association studies called pre-conditioned random forest regression (PRFR). PRFR includes a pre-conditioning step, producing adjusted outcomes, followed by random forest regression modeling. Data was from germline genome-wide SNPs for 668 prostate cancer patients treated with radiotherapy. The cohort was stratified only once, at the outset of the modeling process, into two groups: a training set (2/3 of samples) for modeling and a validation set (1/3 of samples). Post-modeling bioinformatics analysis was conducted to identify biological correlates plausibly associated with the risk of hematuria.

RESULTS

The PRFR method achieved significantly better predictive performance compared to other alternative methods (all p < 0.05). The odds ratio between the high and low risk groups, each of which consisted of 1/3 of samples in the validation set, was 2.87 (p = 0.029), implying a clinically useful level of discrimination. Bioinformatics analysis identified six key proteins encoded by CTNND2, GSK3B, KCNQ2, NEDD4L, PRKAA1, and TXNL1 genes as well as four statistically significant biological process networks previously shown to be associated with the bladder and urinary tract.

CONCLUSION

The risk of hematuria is significantly dependent on common genetic variants. The PRFR algorithm resulted in a stratification of prostate cancer patients at differential risk levels of post-radiotherapy hematuria. Bioinformatics analysis identified important biological processes involved in radiation-induced hematuria.

Improved prediction of drug-induced liver injury literature using natural language processing and machine learning methods

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%.

Predicting cancer relapse following lung stereotactic radiotherapy: an external validation study using real-world evidence

Purpose

For patients receiving lung stereotactic ablative radiotherapy (SABR), evidence suggests that high peritumor density predicts an increased risk of microscopic disease (MDE) and local-regional failure, but only if there is low or heterogenous incidental dose surrounding the tumor (GTV). A data-mining method (Cox-per-radius) has been developed to investigate this dose-density interaction. We apply the method to predict local relapse (LR) and regional failure (RF) in patients with non-small cell lung cancer.

Methods

199 patients treated in a routine setting were collated from a single institution for training, and 76 patients from an external institution for validation. Three density metrics (mean, 90th percentile, standard deviation (SD)) were studied in 1mm annuli between 0.5cm inside and 2cm outside the GTV boundary. Dose SD and fraction of volume receiving less than 30Gy were studied in annuli 0.5-2cm outside the GTV to describe incidental MDE dosage. Heat-maps were created that correlate with changes in LR and RF rates due to the interaction between dose heterogeneity and density at each distance combination. Regions of significant improvement were studied in Cox proportional hazards models, and explored with and without re-fitting in external data. Correlations between the dose component of the interaction and common dose metrics were reported.

Results

Local relapse occurred at a rate of 6.5% in the training cohort, and 18% in the validation cohort, which included larger and more centrally located tumors. High peritumor density in combination with high dose variability (0.5 - 1.6cm) predicts LR. No interactions predicted RF. The LR interaction improved the predictive ability compared to using clinical variables alone (optimism-adjusted C-index; 0.82 vs 0.76). Re-fitting model coefficients in external data confirmed the importance of this interaction (C-index; 0.86 vs 0.76). Dose variability in the 0.5-1.6 cm annular region strongly correlates with heterogeneity inside the target volume (SD; ρ = 0.53 training, ρ = 0.65 validation).

Conclusion

In these real-world cohorts, the combination of relatively high peritumor density and high dose variability predicts increase in LR, but not RF, following lung SABR. This external validation justifies potential use of the model to increase low-dose CTV margins for high-risk patients.

Automatically tracking brain metastases after stereotactic radiosurgery

Background and purpose

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.

Methods

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.

Results

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.

Conclusions

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.

Automated VMAT treatment planning using sequential convex programming: algorithm development and clinical implementation

OBJECTIVE

To develop and clinically implement a fully automated treatment planning system for volumetric modulated arc therapy (VMAT).&#xD;Approach: We solve two constrained optimization problems sequentially. The tumor coverage is maximized at the first step while respecting all maximum/mean dose clinical criteria. The second step further reduces the dose at the surrounding organs-at-risk (OARs) as much as possible. Our algorithm optimizes the machine parameters (leaf positions and monitor units) directly and the resulting mathematical non-convexity is handled using thesequential convex programmingby solving a series of convex approximation problems. We directly integrate two novel convex surrogate metrics to improve plan delivery efficiency and reduce plan complexity by promoting aperture shape regularity and neighboring aperture similarity. The entire workflow is automated using the Eclipse treatment planning system (TPS) application program interface (API) scripting and provided to users as a plug-in, requiring the users to solely provide the contours and their preferred arcs. Our program provides the optimal machine parameters and does not utilize the Eclipse optimization engine, however, it utilizes the Eclipse final dose calculation engine. We have tested our program on 60 patients of different disease sites and prescriptions for stereotactic body radiotherapy (SBRT) (paraspinal (24Gy x 1, 9Gy x 3), oligometastis (9Gy x 3), lung (18Gy x 3, 12Gy x 4)) and retrospectively compared the automated plans with the manual plans used for treatment. The program is currently deployed in our clinic and being used in our daily clinical routine to treat patients. &#xD;Main results: The automated plans found dosimetrically comparable or superior to the manual plans. For paraspinal (24Gy x 1), the automated plans especially improved tumor coverage (the average PTV95% from 96% to 98% and CTV100% from 95% to 97%) and homogeneity (the average PTV maximum dose from 120% to 116%). For other sites/prescriptions, the automated plans especially improved the duty cycle (23%-39.4%).&#xD;Significance: This work proposes a fully automated approach to the mathematically challenging VMAT problem. It also shows how the capabilities of the existing FDA-approved commercial TPS can be enhanced using an in-house developed optimization algorithm that completely replaces the TPS optimization engine. The code and pertained models along with a sample dataset will be released on our ECHO-VMAT GitHub (https://github.com/PortPy-Project/ECHO-VMAT).

Abstract 6541: Geometry of gene expression network reveals potential novel indicator in Ewing sarcoma

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 4657: Topological data analysis reveals pan-cancer immune phenotypes with immune-related survival differences

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 2061: Protein network analysis uncovers a poor-survival subtype in multiple myeloma

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.

Long-term p21 and p53 dynamics regulate the frequency of mitosis events and cell cycle arrest following radiation damage

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.

External validation of pulmonary radiotherapy toxicity models for ultracentral lung tumors

Introduction

Pulmonary toxicity is dose-limiting in stereotactic body radiation therapy (SBRT) for tumors that abut the proximal bronchial tree (PBT), esophagus, or other mediastinal structures. In this work we explored published models of pulmonary toxicity following SBRT for such ultracentral tumors in an independent cohort of patients.

Methods

The PubMed database was searched for pulmonary toxicity models. Identified models were tested in a cohort of patients with ultracentral lung tumors treated between 2008 and 2017 at one large center (N = 88). This cohort included 60 % primary and 40 % metastatic tumors treated to 45 Gy in 5 fractions (fx), 50 Gy in 5 fx, 60 Gy in 8 fx, or 60 Gy in 15 fx prescribed as 100 % dose to PTV.

Results

Seven published NTCP models from two studies were identified. The NTCP models utilized PBT max point dose (Dmax), D0.2 cm³, V65, V100, and V130. Within the independent cohort, the ≥ grade 3 toxicity and grade 5 toxicity rates were 18 % and 7-10 %, respectively, and the Dmax models best described pulmonary toxicity. The Dmax to 0.1 cm³ model was better calibrated and had increased steepness compared to the Dmax model. A re-planning study minimizing PBT 0.1 cm³ to below 122 Gy in EQD2₃ (for a 10 % ≥grade 3 pulmonary toxicity) was demonstrated to be completely feasible in 4/6 patients, and dose to PBT 0.1 cm³ was considerably lowered in all six patients.

Conclusions

Pulmonary toxicity models were identified from two studies and explored within an independent ultracentral lung tumor cohort. A modified Dmax to 0.1 cm³ PBT model displayed the best performance. This model could be utilized as a starting point for rationally constructed airways constraints in ultracentral patients treated with SBRT or hypofractionation.

Nested block self-attention multiple resolution residual network for multiorgan segmentation from CT

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.

Validation of an established deep learning auto-segmentation tool for cardiac substructures in 4D radiotherapy planning scans

•

Cardiotoxicity is a common complication of lung cancer radiotherapy.

•

Segmentation of cardiac substructures is time-consuming and challenging.

•

Deep learning segmentation tools can perform this task in 3D and 4D scans.

•

Performance is high when assessed geometrically, dosimetrically and clinically.

•

Auto-segmentation tools may accelerate clinical workflows and enable research.

Background

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.

Materials and Methods

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.

Results

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.

Conclusions

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.

vWCluster: Vector-valued optimal transport for network based clustering using multi-omics data in breast cancer

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.

aWCluster: A Novel Integrative Network-Based Clustering of Multiomics for Subtype Analysis of Cancer Data

The remarkable growth of multi-platform genomic profiles has led to the challenge of multiomics data integration. In this study, we present a novel network-based multiomics clustering founded on the Wasserstein distance from optimal mass transport. This distance has many important geometric properties making it a suitable choice for application in machine learning and clustering. Our proposed method of aggregating multiomics and Wasserstein distance clustering (aWCluster) is applied to breast carcinoma as well as bladder carcinoma, colorectal adenocarcinoma, renal carcinoma, lung non-small cell adenocarcinoma, and endometrial carcinoma from The Cancer Genome Atlas project. Subtypes were characterized by the concordant effect of mRNA expression, DNA copy number alteration, and DNA methylation of genes and their neighbors in the interaction network. aWCluster successfully clusters all cancer types into classes with significantly different survival rates. Also, a gene ontology enrichment analysis of significant genes in the low survival subgroup of breast cancer leads to the well-known phenomenon of tumor hypoxia and the transcription factor ETS1 whose expression is induced by hypoxia. We believe aWCluster has the potential to discover novel subtypes and biomarkers by accentuating the genes that have concordant multiomics measurements in their interaction network, which are challenging to find without the network inference or with single omics analysis.

Unpaired Cross-Modality Educed Distillation (CMEDL) for Medical Image Segmentation

Accurate and robust segmentation of lung cancers from CT, even those located close to mediastinum, is needed to more accurately plan and deliver radiotherapy and to measure treatment response. Therefore, we developed a new cross-modality educed distillation (CMEDL) approach, using unpaired CT and MRI scans, whereby an informative teacher MRI network guides a student CT network to extract features that signal the difference between foreground and background. Our contribution eliminates two requirements of distillation methods: (i) paired image sets by using an image to image (I2I) translation and (ii) pre-training of the teacher network with a large training set by using concurrent training of all networks. Our framework uses an end-to-end trained unpaired I2I translation, teacher, and student segmentation networks. Architectural flexibility of our framework is demonstrated using 3 segmentation and 2 I2I networks. Networks were trained with 377 CT and 82 T2w MRI from different sets of patients, with independent validation (N = 209 tumors) and testing (N = 609 tumors) datasets. Network design, methods to combine MRI with CT information, distillation learning under informative (MRI to CT), weak (CT to MRI) and equal teacher (MRI to MRI), and ablation tests were performed. Accuracy was measured using Dice similarity (DSC), surface Dice (sDSC), and Hausdorff distance at the 95th percentile (HD95). The CMEDL approach was significantly (p < 0.001) more accurate (DSC of 0.77 vs. 0.73) than non-CMEDL methods with an informative teacher for CT lung tumor, with a weak teacher (DSC of 0.84 vs. 0.81) for MRI lung tumor, and with equal teacher (DSC of 0.90 vs. 0.88) for MRI multi-organ segmentation. CMEDL also reduced inter-rater lung tumor segmentation variabilities.

Pre-treatment immune status predicts disease control in NSCLCs treated with chemoradiation and durvalumab

BACKGROUND

The impact of peripheral blood immune measures and radiation-induced lymphopenia on outcomes in non-small cell lung cancer (NSCLC) patients treated with concurrent chemoradiation (cCRT) and immune check point inhibition (ICI) has yet to be fully defined.

METHODS

Stage III NSCLC patients treated with cCRT and ≥1 dose of durvalumab across a cancer center were examined. Peripheral blood counts were assessed pre-cCRT, during cCRT and at the start of ICI. These measures and risk-scores from two published models estimating radiation dose to immune-bearing organs were tested for association with disease control.

RESULTS

We assessed 113 patients treated with cCRT and a median of 8.5 months of durvalumab. Median PFS was 29 months (95% CI 18-35 months). A lower pre-cCRT ALC (HR: 0.51 (95% CI: 0.32-0.82), p = 0.02) and a higher pre-cCRT ANC (HR: 1.14 (1.06-1.23), p = 0.005) were associated with poor PFS. Neither ALC nadir, ALC at ICI start, ANC at ICI start or the normalized change in ALC from pre-cCRT to nadir were significantly associated with PFS (p = 0.07-0.49). Also, risk scores from the two radiation-dose models were not associated with PFS (p = 0.14, p = 0.21) but were so with the ALC Nadir (p = 0.001, p = 0.002). A higher pre-cCRT NLR was the strongest predictor for PFS (HR: 1.09 (1.05-1.14), p = 0.0001). The 12-month PFS in patients with the bottom vs. top NLR tertile was 84% vs 46% (p = 0.000004).

CONCLUSIONS

Baseline differences in peripheral immune cell populations are associated with disease outcomes in NSCLC patients treated with cCRT and ICI.

Pan-Cancer Prediction of Cell-Line Drug Sensitivity Using Network-Based Methods

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.

Prospectively-validated deep learning model for segmenting swallowing and chewing structures in CT

Objective.Delineating swallowing and chewing structures aids in radiotherapy (RT) treatment planning to limit dysphagia, trismus, and speech dysfunction. We aim to develop an accurate and efficient method to automate this process.Approach.CT scans of 242 head and neck (H&N) cancer patients acquired from 2004 to 2009 at our institution were used to develop auto-segmentation models for the masseters, medial pterygoids, larynx, and pharyngeal constrictor muscle using DeepLabV3+. A cascaded framework was used, wherein models were trained sequentially to spatially constrain each structure group based on prior segmentations. Additionally, an ensemble of models, combining contextual information from axial, coronal, and sagittal views was used to improve segmentation accuracy. Prospective evaluation was conducted by measuring the amount of manual editing required in 91 H&N CT scans acquired February-May 2021.Main results. Medians and inter-quartile ranges of Dice similarity coefficients (DSC) computed on the retrospective testing set (N = 24) were 0.87 (0.85-0.89) for the masseters, 0.80 (0.79-0.81) for the medial pterygoids, 0.81 (0.79-0.84) for the larynx, and 0.69 (0.67-0.71) for the constrictor. Auto-segmentations, when compared to two sets of manual segmentations in 10 randomly selected scans, showed better agreement (DSC) with each observer than inter-observer DSC. Prospective analysis showed most manual modifications needed for clinical use were minor, suggesting auto-contouring could increase clinical efficiency. Trained segmentation models are available for research use upon request viahttps://github.com/cerr/CERR/wiki/Auto-Segmentation-models.Significance.We developed deep learning-based auto-segmentation models for swallowing and chewing structures in CT and demonstrated its potential for use in treatment planning to limit complications post-RT. To the best of our knowledge, this is the only prospectively-validated deep learning-based model for segmenting chewing and swallowing structures in CT. Segmentation models have been made open-source to facilitate reproducibility and multi-institutional research.

Automated and Clinically Optimal Treatment Planning for Cancer Radiotherapy

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 4,000 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.

Geometric network analysis provides prognostic information in patients with high grade serous carcinoma of the ovary treated with immune checkpoint inhibitors

Network analysis methods can potentially quantify cancer aberrations in gene networks without introducing fitted parameters or variable selection. A new network curvature-based method is introduced to provide an integrated measure of variability within cancer gene networks. The method is applied to high-grade serous ovarian cancers (HGSOCs) to predict response to immune checkpoint inhibitors (ICIs) and to rank key genes associated with prognosis. Copy number alterations (CNAs) from targeted and whole-exome sequencing data were extracted for HGSOC patients (n = 45) treated with ICIs. CNAs at a gene level were represented on a protein–protein interaction network to define patient-specific networks with a fixed topology. A version of Ollivier–Ricci curvature was used to identify genes that play a potentially key role in response to immunotherapy and further to stratify patients at high risk of mortality. Overall survival (OS) was defined as the time from the start of ICI treatment to either death or last follow-up. Kaplan–Meier analysis with log-rank test was performed to assess OS between the high and low curvature classified groups. The network curvature analysis stratified patients at high risk of mortality with p = 0.00047 in Kaplan–Meier analysis in HGSOC patients receiving ICI. Genes with high curvature were in accordance with CNAs relevant to ovarian cancer. Network curvature using CNAs has the potential to be a novel predictor for OS in HGSOC patients treated with immunotherapy.

Deep learning auto-segmentation and automated treatment planning for trismus risk reduction in head and neck cancer radiotherapy

Background and Purpose

Reducing trismus in radiotherapy for head and neck cancer (HNC) is important. Automated deep learning (DL) segmentation and automated planning was used to introduce new and rarely segmented masticatory structures to study if trismus risk could be decreased.

Materials and Methods

Auto-segmentation was based on purpose-built DL, and automated planning used our in-house system, ECHO. Treatment plans for ten HNC patients, treated with 2 Gy × 35 fractions, were optimized (ECHO₀). Six manually segmented OARs were replaced with DL auto-segmentations and the plans re-optimized (ECHO₁). In a third set of plans, mean doses for auto-segmented ipsilateral masseter and medial pterygoid (MI_Mean, MPI_Mean), derived from a trismus risk model, were implemented as dose-volume objectives (ECHO₂). Clinical dose-volume criteria were compared between the two scenarios (ECHO₀vs. ECHO₁; ECHO₁vs. ECHO₂; Wilcoxon signed-rank test; significance: p < 0.01).

Results

Small systematic differences were observed between the doses to the six auto-segmented OARs and their manual counterparts (median: ECHO₁ = 6.2 (range: 0.4, 21) Gy vs. ECHO₀ = 6.6 (range: 0.3, 22) Gy; p = 0.007), and the ECHO₁ plans provided improved normal tissue sparing across a larger dose-volume range. Only in the ECHO₂ plans, all patients fulfilled both MI_Mean and MPI_Mean criteria. The population median MI_Mean and MPI_Mean were considerably lower than those suggested by the trismus model (ECHO₀: MI_Mean = 13 Gy vs. ≤42 Gy; MPI_Mean = 29 Gy vs. ≤68 Gy).

Conclusions

Automated treatment planning can efficiently incorporate new structures from DL auto-segmentation, which results in trismus risk sparing without deteriorating treatment plan quality. Auto-planning and deep learning auto-segmentation together provide a powerful platform to further improve treatment planning.

Automatic segmentation of brain metastases using T1 magnetic resonance and computed tomography images

An increasing number of patients with multiple brain metastases are being treated with stereotactic radiosurgery (SRS). Manually identifying and contouring all metastatic lesions is difficult and time-consuming, and a potential source of variability. Hence, we developed a 3D deep learning approach for segmenting brain metastases on MR and CT images. Five-hundred eleven patients treated with SRS were retrospectively identified for this study. Prior to radiotherapy, the patients were imaged with 3D T1 spoiled-gradient MR post-Gd (T1 + C) and contrast-enhanced CT (CECT), which were co-registered by a treatment planner. The gross tumor volume contours, authored by the attending radiation oncologist, were taken as the ground truth. There were 3 ± 4 metastases per patient, with volume up to 57 ml. We produced a multi-stage model that automatically performs brain extraction, followed by detection and segmentation of brain metastases using co-registered T1 + C and CECT. Augmented data from 80% of these patients were used to train modified 3D V-Net convolutional neural networks for this task. We combined a normalized boundary loss function with soft Dice loss to improve the model optimization, and employed gradient accumulation to stabilize the training. The average Dice similarity coefficient (DSC) for brain extraction was 0.975 ± 0.002 (95% CI). The detection sensitivity per metastasis was 90% (329/367), with moderate dependence on metastasis size. Averaged across 102 test patients, our approach had metastasis detection sensitivity 95 ± 3%, 2.4 ± 0.5 false positives, DSC of 0.76 ± 0.03, and 95th-percentile Hausdorff distance of 2.5 ± 0.3 mm (95% CIs). The volumes of automatic and manual segmentations were strongly correlated for metastases of volume up to 20 ml (r=0.97,p<0.001). This work expounds a fully 3D deep learning approach capable of automatically detecting and segmenting brain metastases using co-registered T1 + C and CECT.

Application of Community Detection Algorithm to Investigate the Correlation between Imaging Biomarkers of Tumor Metabolism, Hypoxia, Cellularity, and Perfusion for Precision Radiotherapy in Head and Neck Squamous Cell Carcinomas

The present study aimed to investigate the correlation at pre-treatment (TX) between quantitative metrics derived from multimodality imaging (MMI), including 18F-FDG-PET/CT, 18F-FMISO-PET/CT, DW- and DCE-MRI, using a community detection algorithm (CDA) in head and neck squamous cell carcinoma (HNSCC) patients. Twenty-three HNSCC patients with 27 metastatic lymph nodes underwent a total of 69 MMI exams at pre-TX. Correlations among quantitative metrics derived from FDG-PET/CT (SUL), FMSIO-PET/CT (K1, k3, TBR, and DV), DW-MRI (ADC, IVIM [D, D*, and f]), and FXR DCE-MRI [Ktrans, ve, and τi]) were investigated using the CDA based on a "spin-glass model" coupled with the Spearman's rank, ρ, analysis. Mean MRI T2 weighted tumor volumes and SULmean values were moderately positively correlated (ρ = 0.48, p = 0.01). ADC and D exhibited a moderate negative correlation with SULmean (ρ ≤ -0.42, p < 0.03 for both). K1 and Ktrans were positively correlated (ρ = 0.48, p = 0.01). In contrast, Ktrans and k3max were negatively correlated (ρ = -0.41, p = 0.03). CDA revealed four communities for 16 metrics interconnected with 33 edges in the network. DV, Ktrans, and K1 had 8, 7, and 6 edges in the network, respectively. After validation in a larger population, the CDA approach may aid in identifying useful biomarkers for developing individual patient care in HNSCC.

Radiation-Induced Dyspnea in Lung Cancer Patients Treated with Stereotactic Body Radiation Therapy

Simple Summary

Dyspnea is a common symptomatic side-effect of thoracic radiation therapy. The aim of this study is to build a predictive model of any-grade radiation-induced dyspnea within six months after stereotactic body radiation therapy in patients treated for non-small cell lung cancer. The occurrence of pre-treatment chronic obstructive pulmonary disease and higher relative lungs volume receiving more than 15 Gy as well as heart volume were shown to be risk factors for dyspnea. The obtained results encourage further studies on the topic, which could validate the present organ-based findings and explore the voxel-based landscape of radiation dose sensitivity in the development of dyspnea.

Abstract

In this study, we investigated the prognostic factors for radiation-induced dyspnea after hypo-fractionated radiation therapy (RT) in 106 patients treated with Stereotactic Body RT for Non-Small-Cell Lung Cancer (NSCLC). The median prescription dose was 50 Gy (range: 40–54 Gy), delivered in a median of four fractions (range: 3–12). Dyspnea within six months after SBRT was scored according to CTCAE v.4.0. Biologically Effective Dose (α/β = 3 Gy) volume histograms for lungs and heart were extracted. Dosimetric parameters along with patient-specific and treatment-related factors were analyzed, multivariable logistic regression method with Leave-One-Out (LOO) internal validation applied. Model performance was evaluated by the area under the receiver operating characteristic (ROC) curve (AUC) and calibration plot parameters. Fifty-seven patients (53.8%) out of 106 developed dyspnea of any grade after SBRT (25/57 grade ≥ 2 cases). A three-variable predictive model including patient comorbidity (COPD), heart volume and the relative lungs volume receiving more than 15 Gy was selected. The model displays an encouraging performance given by a training ROC-AUC = 0.71 [95%CI 0.61–0.80] and a LOO-ROC-AUC = 0.64 [95%CI 0.53–0.74]. Further modeling efforts are needed for dyspnea prediction in hypo-fractionated treatments in order to identify patients at high risk for developing lung toxicity more accurately.