Treder Maximilian, Lauermann Jost Lennart, Eter Nicole
Department of Ophthalmology, University of Muenster Medical Center, Domagkstraße 15, 48149, Muenster, Germany.
Graefes Arch Clin Exp Ophthalmol. 2018 Nov;256(11):2053-2060. doi: 10.1007/s00417-018-4098-2. Epub 2018 Aug 8.
To automatically detect and classify geographic atrophy (GA) in fundus autofluorescence (FAF) images using a deep learning algorithm.
In this study, FAF images of patients with GA, a healthy comparable group and a comparable group with other retinal diseases (ORDs) were used to train a multi-layer deep convolutional neural network (DCNN) (1) to detect GA and (2) to differentiate in GA between a diffuse-trickling pattern (dt-GA) and other GA FAF patterns (ndt-GA) in FAF images. 1. For the automated detection of GA in FAF images, two classifiers were built (GA vs. healthy/GA vs. ORD). The DCNN was trained and validated with 400 FAF images in each case (GA 200, healthy 200, or ORD 200). For the subsequent testing, the built classifiers were then tested with 60 untrained FAF images in each case (AMD 30, healthy 30, or ORD 30). Hereby, both classifiers automatically determined a GA probability score and a normal FAF probability score or an ORD probability score. 2. To automatically differentiate between dt-GA and ndt-GA, the DCNN was trained and validated with 200 FAF images (dt-GA 72; ndt-GA 138). Afterwards, the built classifier was tested with 20 untrained FAF images (dt-GA 10; ndt-GA 10) and a dt-GA probability score and an ndt-GA probability score was calculated. For both classifiers, the performance of the training and validation procedure after 500 training steps was measured by determining training accuracy, validation accuracy, and cross entropy.
For the GA classifiers (GA vs. healthy/GA vs. ORD), the achieved training accuracy was 99/98%, the validation accuracy 96/91%, and the cross entropy 0.062/0.100. For the dt-GA classifier, the training accuracy was 99%, the validation accuracy 77%, and the cross entropy 0.166. The mean GA probability score was 0.981 ± 0.048 (GA vs. healthy)/0.972 ± 0.439 (GA vs. ORD) in the GA image group and 0.01 ± 0.016 (healthy)/0.061 ± 0.072 (ORD) in the comparison groups (p < 0.001). The mean dt-GA probability score was 0.807 ± 0.116 in the dt-GA image group and 0.180 ± 0.100 in the ndt-GA image group (p < 0.001).
For the first time, this study describes the use of a deep learning-based algorithm to automatically detect and classify GA in FAF. Hereby, the created classifiers showed excellent results. With further developments, this model may be a tool to predict the individual progression risk of GA and give relevant information for future therapeutic approaches.
使用深度学习算法自动检测和分类眼底自发荧光（FAF）图像中的地图样萎缩（GA）。
在本研究中，使用GA患者、健康对照组和其他视网膜疾病（ORD）对照组的FAF图像训练多层深度卷积神经网络（DCNN），以（1）检测GA，（2）在FAF图像中区分GA的弥漫性细流模式（dt-GA）和其他GA FAF模式（ndt-GA）。1. 对于FAF图像中GA的自动检测，构建了两个分类器（GA与健康对照/GA与ORD）。在每种情况下，使用400张FAF图像（GA 200张、健康对照200张或ORD 200张）对DCNN进行训练和验证。对于后续测试，然后在每种情况下使用60张未训练的FAF图像（年龄相关性黄斑变性30张、健康对照30张或ORD 30张）对构建的分类器进行测试。由此，两个分类器自动确定GA概率分数、正常FAF概率分数或ORD概率分数。2. 为了自动区分dt-GA和ndt-GA，使用200张FAF图像（dt-GA 72张；ndt-GA 138张）对DCNN进行训练和验证。之后，使用20张未训练的FAF图像（dt-GA 10张；ndt-GA 10张）对构建的分类器进行测试，并计算dt-GA概率分数和ndt-GA概率分数。对于两个分类器，在500个训练步骤后，通过确定训练准确率、验证准确率和交叉熵来衡量训练和验证过程的性能。
对于GA分类器（GA与健康对照/GA与ORD），训练准确率达到99%/98%，验证准确率为96%/91%，交叉熵为0.062/0.100。对于dt-GA分类器，训练准确率为99%，验证准确率为77%，交叉熵为0.166。GA图像组中GA概率分数的平均值为0.981±0.048（GA与健康对照）/0.972±0.439（GA与ORD），对照组中为0.01±0.016（健康对照）/0.061±0.072（ORD）（p<0.001）。dt-GA图像组中dt-GA概率分数的平均值为0.807±0.116，ndt-GA图像组中为0.180±0.100（p<0.001）。
本研究首次描述了使用基于深度学习的算法自动检测和分类FAF中的GA。由此，创建的分类器显示出优异的结果。随着进一步发展，该模型可能成为预测GA个体进展风险的工具，并为未来的治疗方法提供相关信息。
