一种基于深度学习的眼科疾病诊断方法

A deep learning based method for ophthalmic disease diagnosis

针对目前逐年上升的眼科疾病尤其是糖尿病性视网膜病(DR)患病率高的问题,设计一种新型眼科疾病诊断方法。该研究采用2019年Kaggle糖尿病性视网膜病检测竞赛数据集,采用先进的深度学习框架,构建多种先进的神经网络模型,经过训练,得到预测结果,最后融合多个模型提高Kappa分数,最终使模型具有较强的健壮性从而解决模型过拟合问题。文中方法包括病情诊断、自动评分分析等功能。实验结果表明,该方法极大地缩短了眼科医生的诊断周期,提高了诊断的准确率和效率,为眼科医生提供了一种便捷、高效的诊断辅助。

Based on the fact of the high prevalence of ophthalmic diseases,especially Diabetic Detinopathy(DR),which is increasing year by year,a new diagnostic method for ophthalmic diseases is designed.This study uses the 2019 Kaggle diabetic retinopathy detection competition data set,adopts the advanced deep learning framework,builds a variety of advanced neural network models.After training,the prediction results are obtained,and finally multiple models are integrated to improve the Kappa score,all owing the final model possessing strong robustness to solve the problem of model over fitting.This method includes functions such as disease diagnosis,automatic score analys is,etc.The experiment results show that this method greatly shortens the diagnosis cycle of ophthalmologists,improves the accuracy and efficiency of diagnosis,and provides a convenient and efficient diagnosis aid for ophthalmologists.