MuDeRN: Multi-category classification of breast histopathological image using deep residual networks
作者：
Highlights：
• MuDeRN is a framework using deep residual network for classifying H&E breast digital.
• MuDeRN classifies patients as benign or cancer with accuracy of 98.77%.
• It classifies benign images into four subtypes with accuracy of adenosis, fibroadenoma, phyllodes tumor, or tubular adenoma.
• It classifies malignant images as ductal carcinoma, lobular carcinoma, mucinous carcinoma, or papillary carcinoma.
• MuDeRN achieved patient-level accuracy of 96.25% for eight-class categorization.
摘要
•MuDeRN is a framework using deep residual network for classifying H&E breast digital.•MuDeRN classifies patients as benign or cancer with accuracy of 98.77%.•It classifies benign images into four subtypes with accuracy of adenosis, fibroadenoma, phyllodes tumor, or tubular adenoma.•It classifies malignant images as ductal carcinoma, lobular carcinoma, mucinous carcinoma, or papillary carcinoma.•MuDeRN achieved patient-level accuracy of 96.25% for eight-class categorization.
论文关键词：Benign breast lesion,Breast cancer,Breast cancer subtypes,Deep learning,Deep residual networks
论文评审过程：Received 28 September 2017, Revised 20 February 2018, Accepted 13 April 2018, Available online 26 April 2018, Version of Record 7 June 2018.
论文官网地址：https://doi.org/10.1016/j.artmed.2018.04.005
