Abstract

Abstract Purpose: To estimate the value of deep learning on magnetic resonance (MR)-base radiomics for T3N1 nasopharyngeal carcinoma (NPC) patients receiving induction chemotherapy (ICT) followed by concurrent chemoradiotherapy (CCRT). Methods: A total of 672 NPC patients with T3N1 disease (training cohort: n = 470; test cohort: n = 202) were enrolled and underwent MRI scan prior to receiving ICT+CCRT. Primary endpoint was disease-free survival (DFS). From each of three sequences of MR images of the primary tumor at baseline, a survival model based on SE-ResNeXt was developed to predict DFS end-to-end. Incorporating three deep learning-based radiomic signatures, a radiomic model was built using multivariable Cox proportional hazard model. Harrell's concordance indices (C-index) were applied to evaluate the predictive power of radiomics nomogram. Results: Three deep learning-based radiomic signatures were significant associated with DFS in the training (C-index: 0.609-0.616) and test (C-index: 0.587-604) cohorts. The radiomic model had the best result for predicting DFS (train cohort: C-index, 0.647, 95% CI, 0.587-0.708, p < 0.001; test cohort: C-index, 0.647, 95% CI, 0.552-0.741, p = 0.002). In addition, the radiomic model also had a significant predictive value for secondary endpoints in the entire dataset (overall survival: C-index, 0.622, p < 0.001; distant metastasis-free survival: C-index, 0.578, p = 0.044; locoregional relapse-free survival: C-index, 0.648, p < 0.001). Conclusion: Deep learning MR-based radiomic signatures may serve as a powerful tool for prognosis prediction in T3N1 nasopharyngeal carcinoma patients receiving ICT + CCRT treatment. Citation Format: Lianzhen Zhong, Di Dong, Linglong Tang, Shuyan Han, Jie Tian. Deep learning-based prognosis prediction in T3N1 nasopharyngeal carcinoma patients treated with induction chemotherapy followed by concurrent chemoradiotherapy [abstract]. In: Proceedings of the Annual Meeting of the American Association for Cancer Research 2020; 2020 Apr 27-28 and Jun 22-24. Philadelphia (PA): AACR; Cancer Res 2020;80(16 Suppl):Abstract nr 5430.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call