Department of Vision Science, University of California, Berkeley, Berkeley, California; International Computer Science Institute, Berkeley, California; Department of Ophthalmology & Vision Science, University of California, Davis, Sacramento, California.
International Computer Science Institute, Berkeley, California.
Ophthalmol Retina. 2022 Feb;6(2):116-129. doi: 10.1016/j.oret.2021.06.010. Epub 2021 Jul 2.
Diseases such as age-related macular degeneration (AMD) are classified based on human rubrics that are prone to bias. Supervised neural networks trained using human-generated labels require labor-intensive annotations and are restricted to specific trained tasks. Here, we trained a self-supervised deep learning network using unlabeled fundus images, enabling data-driven feature classification of AMD severity and discovery of ocular phenotypes.
Development of a self-supervised training pipeline to evaluate fundus photographs from the Age-Related Eye Disease Study (AREDS).
One hundred thousand eight hundred forty-eight human-graded fundus images from 4757 AREDS participants between 55 and 80 years of age.
We trained a deep neural network with self-supervised Non-Parametric Instance Discrimination (NPID) using AREDS fundus images without labels then evaluated its performance in grading AMD severity using 2-step, 4-step, and 9-step classification schemes using a supervised classifier. We compared balanced and unbalanced accuracies of NPID against supervised-trained networks and ophthalmologists, explored network behavior using hierarchical learning of image subsets and spherical k-means clustering of feature vectors, then searched for ocular features that can be identified without labels.
Accuracy and kappa statistics.
NPID demonstrated versatility across different AMD classification schemes without re-training and achieved balanced accuracies comparable with those of supervised-trained networks or human ophthalmologists in classifying advanced AMD (82% vs. 81-92% or 89%), referable AMD (87% vs. 90-92% or 96%), or on the 4-step AMD severity scale (65% vs. 63-75% or 67%), despite never directly using these labels during self-supervised feature learning. Drusen area drove network predictions on the 4-step scale, while depigmentation and geographic atrophy (GA) areas correlated with advanced AMD classes. Self-supervised learning revealed grader-mislabeled images and susceptibility of some classes within more granular AMD scales to misclassification by both ophthalmologists and neural networks. Importantly, self-supervised learning enabled data-driven discovery of AMD features such as GA and other ocular phenotypes of the choroid (e.g., tessellated or blonde fundi), vitreous (e.g., asteroid hyalosis), and lens (e.g., nuclear cataracts) that were not predefined by human labels.
Self-supervised learning enables AMD severity grading comparable with that of ophthalmologists and supervised networks, reveals biases of human-defined AMD classification systems, and allows unbiased, data-driven discovery of AMD and non-AMD ocular phenotypes.
年龄相关性黄斑变性（AMD）等疾病是基于易产生偏差的人类分类法进行分类的。使用人类生成的标签进行监督的神经网络训练需要大量的注释，并且仅限于特定的训练任务。在这里，我们使用未标记的眼底图像训练了一个自我监督的深度学习网络，从而实现了 AMD 严重程度的基于数据的特征分类和眼部表型的发现。
开发一个自我监督的培训管道，以评估来自年龄相关性眼病研究（AREDS）的眼底照片。
4757 名年龄在 55 至 80 岁之间的 AREDS 参与者的 10848 张人类分级眼底图像。
我们使用自我监督的非参数实例鉴别（NPID）训练深度神经网络，该网络使用无标签的 AREDS 眼底图像，然后使用两步、四步和九步分类方案使用监督分类器评估其在 AMD 严重程度分级中的性能。我们比较了 NPID 与监督训练网络和眼科医生的平衡和不平衡准确率，使用图像子集的分层学习和特征向量的球形 K-均值聚类探索网络行为，然后搜索无需标签即可识别的眼部特征。
准确率和 Kappa 统计。
NPID 在不同的 AMD 分类方案中表现出多功能性，无需重新训练即可实现平衡准确率，与监督训练网络或人类眼科医生在分类晚期 AMD（82%比 81-92%或 89%）、可归因 AMD（87%比 90-92%或 96%）或 4 步 AMD 严重程度分级（65%比 63-75%或 67%）方面的准确率相当，尽管在自我监督的特征学习过程中从未直接使用这些标签。NPID 主要根据玻璃膜疣面积对 4 步分级进行预测，而脱色素和地图状萎缩（GA）面积与晚期 AMD 类别相关。自我监督学习揭示了分级员错误标记的图像，以及一些更精细的 AMD 分级中某些类别的易混淆性，包括眼科医生和神经网络。重要的是，自我监督学习使数据驱动的 AMD 特征发现成为可能，例如 GA 和脉络膜（如格子状或金发眼底）、玻璃体（如星状白内障）和晶状体（如核性白内障）的其他眼部表型，这些特征并非由人类标签预先定义。
自我监督学习可以实现与眼科医生和监督网络相当的 AMD 严重程度分级，揭示了人类定义的 AMD 分类系统的偏见，并允许对 AMD 和非 AMD 眼部表型进行无偏见、基于数据的发现。
