糖尿病眼底视网膜病变辅助诊断系统设计

Design of Auxiliary Diagnosis System for Diabetic Fundus Retinopathy

糖尿病视网膜病变筛查对于患者预防致盲有重要意义。基于计算机的辅助诊断系统可快速而精确地处理分析图像数据,为医生提供客观的数字化诊断依据。然而,当病理图像分辨率较低或类别之间差异小时,分类精度仍有待提升。文章设计融合注意力机制(CBAM Convolutional Block Attention Module)的ResNet深度学习网络模型应用于糖尿病视网膜病变筛查。设计微信小程序方便用户查看检测结果,实现糖尿病眼底视网膜病变辅助系统。实验结果表明,本文所设计的系统对糖尿病视网膜病变的识别准确率可达94.72%,灵敏度为92.67%,特异度为99.83%,可精确检测糖尿病视网膜病变,且为患者提供了便捷的客户端平台。

Screening for diabetic retinopathy is important for preventing blindness.The comput-er-aided diagnosis system can process and analyze image data quickly and accurately,and provide objective digital diagnosis basis for doctors.However,when the resolution of pathological ima-ges is low or the difference betwecn catcgorics is small,the classification accuracy still nceds to be improved.In this paper,a CBAM Convolutional Block Attention Module based ResNet deep learning network model is designed and applied to the screening of diabetic retinopathy.Wechat mini program was designed to facilitate users to view the test results and realize an auxiliary sys-tem for diabetic fundus retinopathy.The experimental results show that the system designed in this paper can accurately detect diabetic retinopathy with the recognition accuracy of 94.72%,sensitivity of 92.67%,specificity of 99.83%,and provide a convenient client platform for pa-tients.