Abstract

Info

Title:: Glomerular pathological image classification algorithm based on convolutional neural network

Author(s):: SUN Xiaohan1; KONG Xiangyong1; WU Ying2; WANG Ping3; CAI Jian1; PENG Ruiyang1; WANG Yuze1; 1. School of Health Science and Engineering, University of Shanghai for Science and Technology, Shanghai 200093, China 2. Department of Pathology, Childrens Hospital Affiliated to Shanghai Jiao Tong University School of Medicine, Shanghai 200040, China 3. Department of Nephrology and Rheumatology, Childrens Hospital Affiliated to Shanghai Jiao Tong University School of Medicine, Shanghai 200040, China

Keywords:: convolutional neural network SE-ResNet image classification glomerulus pathological image

Abstract:: Abstract: The automatic classification of glomeruli in pathological sections is the key to diagnose the degree and type of renal lesions. A complete glomerular classification framework based on convolutional neural network is designed for realizing glomerular classification. SE-Resnet is used as the image classification model to change the convolutional layer in the original module into convolution blocks with a smaller number of parameters, thereby reducing network parameters on the premise of ensuring network performance. The experimental results show that compared with other classification algorithms, the proposed algorithm has the optimal performance and achieves 96.93% accuracy in the classification of mesangial proliferative glomerulonephritis, crescentic glomerulonephritis, focal segmental glomerulosclerosis and normal glomeruli, indicating that the classification algorithm can better identify glomerular lesions.