Research on Classification of Hepatic Steatosis Based on Deep Learning
LI Xinghui1，ZHANG Zheng1，ZHANG Lei2，HE Fuquan2，WANG Feng1
1. School of Biological Sciences and Medical Engineering, Southeast University, Nanjing Jiangsu 210096, China丨
2. Department of Radiology, Shanghai General Hospital, Shanghai 201620, China
Abstract：Objective To research classification of hepatic steatosis combined with deep learning and transfer learning, the research
provided an intelligent and non-invasive method for classification of hepatic steatosis. Methods This retrospective study included 50
patients who underwent mDixon imaging sequence scan of upper abdomen in Shanghai General Hospital, from June to July, 2018.
Region of interest from 50 MR images was processed for data enhancement and unbalanced data processing. Results The AUC
(area under the ROC curve) of DenseNet model in validation cohort was 0.83. The accuracy, sensitivity and specificity were 84.96%,
94.55%, and 79.82%. Conclusion Classification of hepatic steatosis based on deep learning method shows advantages of performance
and automation. Correlation analysis provides a new method to analysis classification of hepatic steatosis and relevant factors.
李星辉1，张政1，张蕾2，何付权2，汪丰1. 基于深度学习的肝脂肪变性分级研究[J]. 中国医疗设备, 2021, 36(6): 45-49.
LI Xinghui1，ZHANG Zheng1，ZHANG Lei2，HE Fuquan2，WANG Feng1. Research on Classification of Hepatic Steatosis Based on Deep Learning. China Medical Devices, 2021, 36(6): 45-49.
