医学信息2025，Vol.38Issue(22)：26-31,6.DOI:10.3969/j.issn.1006-1959.2025.22.004
基于眼底图像诊断的神经网络分类模型研究
Research on Neural Network Classification Model Based on Fundus Image Diagnosis
摘要
Abstract
With the development and progress of computer-aided diagnosis system in the medical field,fundus images carry rich clinical information,which can be well combined with computer-aided diagnosis technology,so as to assist clinicians in judging ophthalmic diseases and reduce the intensity and burden of doctors'work.In this study,1000 fundus images were collected as validation data sets.From the perspective of clinical fundus image diagnosis,a neural network was built with InceptionV3 structure as the core,which strengthened the processing of over-fitting of the model.Dropout technology was used in the model construction.In addition,the scheme of fundus image data augmentation was improved,and more correct clinical information was input into the model for training to ensure that more than 80%of the fundus images in the data set contain relevant clinical features.The precision of the training model was 94.18%.In the validation data set collected by the hospital,the validation precision can reach 90.85%.The overall accuracy of the model was as expected.Compared with the prediction results of other experiments,the accuracy of the model was relatively high,the performance was relatively stable,and the clinical disease classification information of the fundus image could be distinguished.关键词
眼底图像/图像识别/卷积神经网络Key words
Fundus image/Image recognition/Convolutional neural network分类
信息技术与安全科学引用本文复制引用
丁廷波..基于眼底图像诊断的神经网络分类模型研究[J].医学信息,2025,38(22):26-31,6.基金项目
2021年泸州市科技计划项目(编号:2021-RCM-111) （编号:2021-RCM-111）
