首都医科大学学报 ›› 2024, Vol. 45 ›› Issue (4): 678-687.doi: 10.3969/j.issn.1006-7795.2024.04.018

• 基础研究 • 上一篇    下一篇

GhostNet 轻量级网络在糖尿病视网膜病变诊断中的应用价值

朱小红1 , 张  云2,  刘美玲3,  曹  凯4*   

  1. 1.北京西城区妇幼保健院内科,北京 100054; 2.建国门社区卫生服务中心全科,北京 100005; 3.大红门社区卫生服务中心保健科,北京  100075; 4.北京市眼科研究所,首都医科大学附属北京同仁医院,北京 100005
  • 收稿日期:2023-09-11 出版日期:2024-08-21 发布日期:2024-07-08
  • 通讯作者: 曹 凯 E-mail:caozhi@ccmu.edu.cn
  • 基金资助:
    北京市医院管理中心“青苗计划”专项(QMS20210215)。

Application value of GhostNet lightweight network in diagnosis of diabetic retinopathy

Zhu  Xiaohong1, Zhang Yun2, Liu Meiling3, Cao Kai4*   

  1. 1.Internal Medicine, Xicheng District Maternal and Child Health Hospital, Beijing 100054,China; 2.General Practice, Jianguomen Community Health Service Center, Beijing 100005,China;3.Department of Health Care, Dahongmen Community Health Service Center, Beijing  100075,China; 4.Beijing Institute of Ophthalmology, Beijing Tongren Hospital, Capital Medical University, Beijing 100005,China
  • Received:2023-09-11 Online:2024-08-21 Published:2024-07-08
  • Supported by:
    This study was supported by Beijing Municipal Administration of Hospitals' Youth Programme (QMS20210215).

摘要: 目的  基于眼底彩照,分别应用经典卷积神经网络DenseNet121和轻量级网络GhostNet训练糖尿病视网膜病变(diabetic retinopathy, DR)的诊断模型(将DR和正常眼底做区分)和鉴别诊断(将DR和其他眼底病做区分)模型,评价基于轻量级网络GhostNet的DR诊断模型的应用价值。方法  收集大样本的眼底彩照29 535张(含DR 9 883张、正常眼底2 000张、用于做鉴别诊断的其他致盲性眼底病17 652张)。分别采用经典卷积神经网络DenseNet121和轻量级网络GhostNet建模,并借助迁移学习做模型训练。采用受试者工作特征(receiver operating characteristic, ROC)曲线及其曲线下面积(area under the curve, AUC)、灵敏度、特异度、准确率评价模型性能。结果  与基于DenseNet121的模型相比,基于GhostNet的模型对单张眼底照的诊断时间缩短了60.3%。在DR的诊断方面,基于GhostNet的模型的AUC值、灵敏度、特异度、准确率分别为0.911、0.888、0.934、91.3%,基于DenseNet121的模型的AUC值、灵敏度、特异度、准确率分别为0.954、0.921、0.986、95.5%。在DR与其他眼底病的鉴别诊断方面,基于GhostNet的模型的AUC值、灵敏度、特异度、准确率分别为0.862、0.856、0.901、87.8%;基于DenseNet121的模型的AUC值、灵敏度、特异度、准确率分别为0.899、0.871、0.935、90.2%。结论  基于GhostNet轻量级神经网络构建的DR诊断模型和鉴别诊断模型,其诊断效率较经典模型DenseNet121有显著提升,并且模型兼具较高的准确率。对于社区医院等缺乏眼科医师且设备性能不高的基层医疗机构,可考虑应用该技术开展DR的初筛。

关键词: 糖尿病视网膜病变, 轻量级神经网络模型, 诊断, 筛查, 社区

Abstract: Objective  To evaluate the value of GhostNet lightweight network based on fundus color imaging in the diagnosis of diabetic retinopathy (DR), and the value of differential diagnosis between DR and other fundus diseases compared with classical convolutional neural network DenseNet121.Methods  Totally 29 535 color fundus photographs (including 9 883 DR, 2 000 normal fundus and 17 652 other blinding fundus diseases for differential diagnosis) were collected from large samples. Classical convolutional neural network (DenseNet121) and lightweight network (GhostNet) were used to train the model by transfer learning. Receiver operating characteristic (ROC)curve, sensitivity, specificity, positive predictive value, negative predictive value and area under the curve (AUC) were used to evaluate the performance of the two models. Results  Compared with model based on DenseNet121, the diagnostic time of single fundus photo by GhostNet model was shortened by 60.3%. In the diagnosis of DR, AUC value, sensitivity and specificity, accuracy reached 0.911, 0.888, 0.934, and 91.3% respectively. It is slightly lower than 0.954, 0.921, 0.986, and 95.5% of model based on DenseNet121. In the differential diagnosis of DR and other fundus diseases, the AUC value, sensitivity, specificity, accuracy of model based on GhostNet were 0.862, 0.856, 0.901, and 87.8% respectively. For model based on DenseNet121, the values were 0.899, 0.871, 0.935, and 90.2%, respectively. Conclusions  The operation speed of DR diagnosis model and differential diagnosis model based on GhostNet lightweight neural network is significantly faster than that of classical model DenseNet121, and the accuracy is satisfying. For primary medical institutions with lack of ophthalmologists and low equipment performance, such as community hospitals, we can consider using this technology to carry out primary screening of DR.

Key words: diabetic retinopathy, lightweight neural network model, diagnosis, screening, community

中图分类号: