Search

Previously, PET image biomarkers (PET-IBMs) - the 90thpercentile standardized uptake value (P90-SUV) and the Mean SUV (Mean-SUV) ofthe contralateral parotid gland (cPG) - were identified as predictors for late-xerostomiafollowing head and neck cancer (HNC) radiotherapy. The aim of the current study wasto assess in an independent validation cohort whether these pre-treatment PET-IBMscan improve late-xerostomia prediction compared to the prediction with baselinexerostomia and mean cPG dose alone.The prediction endpoint was patient-rated moderate-to-severe xerostomia at 12 months after radiotherapy. The PET-IBMs were extracted frompre-treatment 18 F-FDG PET images. The performance of the model (base model)with baseline xerostomia and mean cPG dose alone and models with additionallyP90-SUV or Mean-SUV were tested in the current independent validation cohort.Specifically, model discrimination (area under the curve: AUC) and calibration(calibration plot) were evaluated.The current validation cohort consisted of 137 patients of which 40% developed moderate-to-severe xerostomia at 12 months. Both the PET-P90 model (AUC PET-P90 =0.71) and the PET-Mean model (AUC PET-Mean =0.70) performed well in the current validation cohort. Moreover, their performance was improved compared to the base model (AUC base =0.68). The calibration plots showed a good fit of the prediction to the actual rates for all tested models.PET-IBMs showed an improved prediction of late-xerostomia when added to the base model in this validation cohort. This contributed to the published hypothesis that PET-IBMs include individualized information and can serve as a pre- treatment risk factor for late-xerostomia.

PURPOSE/OBJECTIVE(S) Post-treatment symptoms are a focal point of follow-up visits for head and neck cancer patients. While symptoms such as dysphagia and shortness of breath may empirically guide earlier detection of disease recurrence, their association with patient outcomes is less well studied, especially for HPV-positive disease. This study aimed to investigate the association between patient reported dysphagia and shortness of breath at 3-to-6 months following radiotherapy (RT) and overall survival (OS). MATERIALS/METHODS Data were collected from oropharyngeal cancer patients that received RT with curative intent from 2015-2019. Included patients had self-reported scores for shortness of breath, choking/coughing ("choke"), and difficulty swallowing/chewing ("swallow"), which were surveyed via the MD Anderson Symptom Inventory-Head and Neck Module (MDASI-HN) questionnaire at 3-to-6 months after RT. Symptom severity was scored from 0-10. Significant predictors of OS were analyzed via Cox regression. Symptom scores were also discretized for Kaplan-Meier (KM) analysis, with more severe symptoms being 6+ for "swallow" and "choke" and 2+ for shortness of breath. RESULTS Symptom scores were collected at 3-to-6 months from 470 patients. The majority (91.3%) were HPV-positive. Median follow-up time was 31.7 months (IQR: 21.9-42.1). Univariable Cox regression (Table 1) showed significant associations between OS and MDASI scores for shortness of breath, "choke," and "swallow." Furthermore, a composite variable integrating scores of all three symptoms had the best predictive value for OS (c-index = 0.748). Symptom severity stratified patients based on OS for all three symptoms, with more severe symptoms predicting worse OS on KM analysis (P < 0.001 for all symptoms). On subset analysis more severe symptoms remained significantly associated with worse OS among HPV-positive patients (P < 0.01 for all symptoms). CONCLUSION Quantitative early patient-reported measures of dysphagia and shortness of breath are significant predictors of OS in head and neck cancer, in particular for HPV-positive disease. Integrating quantitative symptom surveys into post-treatment surveillance may be beneficial for individualizing follow-up strategies.

Purpose/Objective(s) This study aims to identify radiomic features extracted from contrast-enhanced CT scans that differentiate osteoradionecrosis (ORN) from normal mandibular bone in head and neck cancer patients treated with radiotherapy. Materials/Methods Contrast-enhanced CT images were collected for patients with confirmed ORN diagnosis at MD Anderson Cancer Center between 2008 and 2018. The ORN region of interest (ROI) was segmented manually in each image. The control ROIs of the contralateral health mandible were generated by a Python script then adjusted manually in each image. An open-source software was then used to extract the radiomic features from both ORN and control ROIs after the application of intrinsic filters. The pairwise correlation filter was used to remove radiomic features whose pairwise correlation was ≥0.99. Filter algorithms were then used to further reduce the number of radiomic features. After that, wrapper and embedded methods were applied on the resulting radiomic features. Finally, Gini importance and Recursive Feature Elimination (RFE) were used to select the final radiomic features for the predictive model. The support vector machine (SVM) with linear kernel was used for the binary classification of ORN and normal mandibular bone. The performance of the model was evaluated using the Area Under Curve (AUC). Results 150 patients with radiologically established ORN were included. The mean age was 62.3 years (range 27-82). The mean time between the end of RT and ORN detection was 32.6 months. A total of Initial 1316 radiomic features were considered. The pairwise correlation omitted 432 features with a correlation ≥ 0.99. The RFE based on the Gini index selected 5 radiomics features in our HNC cohort. We validated this binary classification model using 5-fold cross-validation. During this validation, the range of AUC was (0.84–0.95) & the average AUC was 0.90. This AUC range reflect the high performance of the final classifier in the differentiation between ORN and normal mandible using CECT images. Conclusion Radiomic features were successfully used to build a model to discriminate ORN and normal mandibular bone in head and neck cancer patients. Our statistical model achieved satisfying prediction accuracy and can be potentially used for ORN prediction purpose upon external validation.

Purpose/Objective(s): Image biomarkers of the contoured Gross Tumorvolume (GTV) in computed tomography (CT) images could be related to tumor biology and prognosis. This study aimed to develop and validate a model to predict overall survival (OS) of nasopharyngeal cancer (NPC)patients, based on tumor Image biomarkers in addition to clinical parameters.Such a model could be used as a pre-treatment tool to support clinical decision-making.Materials/Methods: This retrospective study was composed of 296 NPC patients treated with (chemo-) radiation therapy. In total, 134 Image biomarkers representing tumor image intensity, shape and texture and clinical parameters (TNM staging, WHO score, dose parameters, age and gender)were analyzed. Using forward selection Cox proportional hazards regression analysis was performed to create a survival model. The concordance index (c-index) was determined to assess the model’s discriminative power.We compared the Kaplan-Meier survival curves between patients with predicted values below and above the median with a log-rank test. Subsequently,the model was externally validated in 310 head and neck cancer(HN) patients (mainly or opharyngeal and laryngeal).Results: Firstly, a multi variable model was developed based on clinical features only (c-index 0.67). Four independent prognostic factors were identified: age, T stage, N stage and D98 (minimum dose in 98% of theGTV). Secondly, another multivariable model was developed based on Image biomarkers only (c-index 0.68). Two variables were identified as most important prognostic factors: “volume density,” describing the compactness of the tumor, and “Run Length Non-Uniformity,” a measurefor intratumor heterogeneity. Interestingly, when Image biomarkers and clinical features were combined in the model, the c-index of the model improved to 0.73, which was significantly better than that of models with Image biomarkers or clinical features alone (Table 1). The performance of the model was also good (c-index Z 0.70) when externally validated inHN patients. The Kaplan-Meier survival curves were significantly different between high and low predicted values both in the training group (the probability of OS after 3 years were 91% and 73%, P