Radiation and Cancer Physics
SS 18 - Physics 4 - Imaging for Response Assessment I
143 - Radiomics Analysis of Skeletal Muscle Predicts Grade 3 or Higher Acute Gastrointestinal Toxicity in Rectal Cancer Patients Receiving Neoadjuvant Chemoradiotherapy
Tuesday, October 23
8:35 AM - 8:45 AM
Location: Lila Cockrell Theatre
Radiomics Analysis of Skeletal Muscle Predicts Grade 3 or Higher Acute Gastrointestinal Toxicity in Rectal Cancer Patients Receiving Neoadjuvant Chemoradiotherapy
F. Xia, W. Yang, Y. Yuan, J. Wang, W. Hu, J. Zhu, and Z. Zhang; Department of Radiation Oncology, Fudan University Shanghai Cancer Center, Shanghai, China
Purpose/Objective(s): Recent retrospective study showed that measurement of skeletal muscle and body composition can serve as a surrogate for patients' nutritional status which associated with treatment outcome and complications. In this study we investigate radiomics features from skeletal muscle on computed-tomography (CT) scans obtained prior to treatment to predict severe toxicities in rectal cancer patients treated with concurrent chemoradiotherapy.
Materials/Methods: Seventy-six patients with locally advanced rectal cancer underwent neoadjuvant pelvic radiotherapy with concurrent weekly administration of capecitabine and irinotecan between June and October, 2017 were retrospectively included. The skeletal muscles at the level of the third lumber vertebral were contoured on images of simulation CT scan and the radiomics features were extracted. Clinical information were recorded including age, gender, body mass index (BMI), cycles of weekly chemotherapy, hematologic and gastrointestinal toxicities (diarrhea and/or proctitis) assessed by RTOG acute toxicity criteria. Binomial logistic regression model was built for grade 3 or higher toxicities using clinical factors and pretreatment radiomics features. Concordance (c) index were used to test predictive accuracy of the model.
Results: The median age was 57 years and 23 patients were women. The median BMI was 22.8 kg/m2 and 51 (67.1%) patients completed at least 4 cycles of weekly chemotherapy. 18 patients developed grade 3 or higher gastrointestinal toxicity (23.6%) and 26 patients developed grade 3 or higher hematologic toxicity (34.2%). A total of 121 radiomics features were calculated using in-house radiomics software after segmentation. Eleven radiomics features were associated with grade 3 or higher gastrointestinal toxicity and included in binomial logistic regression. The C-index of the prediction model was 0.9167. None of the conventional imaging measurements of muscle size (i.e, volume, area and muscle index) or clinical factors (i.e, age, sex, BMI and weekly chemotherapy cycles) were predictive for grade 3+ toxicities.
Conclusion: This study showed that radiomics features of skeletal muscle can predict severe acute gastrointestinal toxicity in rectal cancer patients receiving neoadjuvant chemoradiotherapy, while conventional measurements and clinical factors were not predictive in this cohort. Thus radiomics features may be a useful biomarker for treatment outcomes that conventional parameters cannot predict.
Author Disclosure: F. Xia: None. W. Yang: None. W. Hu: None. Z. Zhang: Independent Contractor; Varian Medical System.