Abstract

Info

Title:: Predictive model for survival in gallbladder cancer patients based on 3D-ResNet deep image features

Author(s):: YIN Ziming1; DONG Dongmin1; CHEN Tao2; 1. School of Medical Instrument and Food Engineering, University of Shanghai for Science and Technology, Shanghai 200093, China 2. Department of Biliary-Pancreatic Surgery, Renji Hospital Affiliated to Shanghai Jiaotong University School of Medicine, Shanghai 200127, China

Keywords:: Keywords: deep learning three-dimensional convolutional neural network gallbladder cancer survival model prognosis

Abstract:: Abstract: Establishing an accurate individualized predictive model for survival in gallbladder cancer patients and finding new prognostic factors for gallbladder cancer are of great significance for prognosis evaluation, treatment model selection, surgical patient screening, postoperative adjuvant treatment plan determination, and rational utilization of medical resources. A method to build a predictive model for survival in gallbladder cancer patients based on deep image features extracted by 3D-ResNet is proposed in the study. The deep features of the patients CT image are automatically extracted through transfer learning and 3D-ResNet training, and the extracted deep image features are used to establish a predictive model for survival in gallbladder cancer patients through the Cox proportional hazard regression model. The experimental results show that the C index of prognostic factors obtained based on deep image features is 0.734 for predicting survival in gallbladder cancer patients, and that the AUC of 1-year, 3-year and 5-year survival rates predicted by deep image features reaches 0.833, 0.791 and 0.813, respectively. The proposed method has a good indication for predicting the prognosis of gallbladder cancer.