摘要
目的 针对糖尿病病患视网膜病变的智能筛查模型构建中数据不足的问题,提出基于迁移学习的解决方案。方法 对少量不同患病程度的视网膜眼底图像做特征强化后,结合ImageNet数据构建的模型进行眼底图像数据的降维,并构建视网膜病变图像的分类模型。结果 最终模型的测试集上准确率达到86.4%。结论 视网膜眼底图像可作为眼部病症的重要判断依据,基于智能模型的检测可代替人工对患者病情进行检查,对开展糖尿病性视网膜病变筛查工作具有重要的现实意义。
Objective To solve the problem of insufficient data of intelligent screening model for diabetic retinopathy, a solution based on transfer learning was proposed. Methods After feature enhancement of a small number of retinal fundus images with different degrees of disease, the fundus image data were reduced by combining the model constructed by ImageNet data, and the classification model of retinopathy images was constructed. Results The Method work and the accuracy of the final model on the test set was 86.4%. Conclusion Retinal fundus images can be used as an important basis for judging ocular diseases. Intelligent model-based detection can replace manual examination of patients’ disease course, and it has very important practical significance to carry out the screening work of diabetic retinopathy.
作者
段以卓
戴妍晴
贺惠新
DUAN Yizhuo;DAI Yanqing;HE Huixin(Harbin Medical University,Harbin 150081,Heilongjiang,China;College of Computer Science and Technology,Huaqiao University,Xiamen 361021.,Fujian,China)
出处
《中国分子心脏病学杂志》
CAS
2022年第3期4699-4705,共7页
Molecular Cardiology of China
关键词
糖尿病
视网膜病变
迁移学习
智能筛查
深度学习
Diabetes
Retinopathy
Transfer learning
Intelligent screening
Deep learning
