Multi-class classification of pathological myopia based on fundus photography  

作　　者：Jiaqing Zhao Guogang Cao Jiangnan He Cuixia Dai 

机构地区：[1]Shanghai Institute of Technology,No.100 Haiquan Road,Fengxian District,Shanghai,P.R.China [2]Shanghai Eye Diseases Prevention&Treatment Center,Shanghai Eye Hospital,School of Medicine,Tongji University,Shanghai 200331,P.R.China

出　　处：《Journal of Innovative Optical Health Sciences》2024年第6期125-136,共12页创新光学健康科学杂志（英文）

基　　金：supported by the Natural National Science Foundation of China(62175156);the Science and technology innovation project of Shanghai Science and Technology Commission(22S31903000);Collaborative Innovation Project of Shanghai Institute of Technology(XTCX2022-27)。

摘　　要：Pathological myopia(PM)is a severe ocular disease leading to blindness.As a traditional noninvasive diagnostic method,fundus color photography(FCP)is widely used in detecting PM due to its highfidelity and precision.However,manual examination of fundus photographs for PM is time-consuming and prone to high error rates.Existing automated detection technologies have yet to study the detailed classification in diagnosing different stages of PM lesions.In this paper,we proposed an intelligent system which utilized Resnet101 technology to multi-categorically diagnose PM by classifying FCPs with different stages of lesions.The system subdivided different stages of PM into eight subcategories,aiming to enhance the precision and efficiency of the diagnostic process.It achieved an average accuracy rate of 98.86%in detection of PM,with an area under the curve(AUC)of 98.96%.For the eight subcategories of PM,the detection accuracy reached 99.63%,with an AUC of 99.98%.Compared with other widely used multi-class models such as VGG16,Vision Transformer(VIT),EfficientNet,this system demonstrates higher accuracy and AUC.This artificial intelligence system is designed to be easily integrated into existing clinical diagnostic tools,providing an efficient solution for large-scale PM screening.

关 键 词：Fundus color photography pathological myopia deep learning Resnet101. 

分 类 号：R778.11[医药卫生—眼科]