一种用于视网膜静脉阻塞分类和病变检测的混合卷积神经网络

A Mixed Convolutional Neural Networks for Retinal Vein Occlusion Classification and Lesions Detection

视网膜静脉阻塞(Retinal Vein Occlusion,RVO)是最常见的视网膜血管疾病之一。提出一种用于视网膜静脉阻塞分类和病灶识别的混合卷积神经网络,将压缩优化后的VGG卷积神经网络(Convolutional Neural Network,CNN)结合类激活层(Class Activation Mapping,CAM)网络构建复合模型,完成眼底彩照上的RVO分类和无监督病变区域识别。通过由山西省眼科医院三位高级眼科医生收集和标记的2962个真实数据进行仿真实验,模型的分类准确率、召回率和F1值均在0.95以上。

Retinal Vein Occlusion(RVO)is one of the most common retinal vascular diseases.In this paper,a mixed convolutional neural network for classification and lesion recognition of RVO is proposed.The compression-optimized VGG is combined with the CAM network to construct a composite model.The proposed model can be used for automated RVO classification and lesions area identification on the fundus color photograph.A simulation experiment was carried out with 2,962 real data collected and marked by three senior ophthalmologists of Shanxi Provincial Eye Hospital.As a result,the classification accuracy,recall rate and F1 value of the model were all above 0.95.