Download This Paper
Deep Learning Achieves Perfect Anomaly Detection on 108,308 Retinal Images Including Unlearned Diseases
28 Pages Posted: 7 May 2020 Publication Status: Review Complete
More...Abstract
Optical coherence tomography (OCT) scanning is useful in detecting various retinal diseases. However, there are not enough ophthalmologists who can diagnose retinal OCT images. To provide OCT screening inexpensively and extensively, an automated diagnosis system is indispensable. Although many machine learning techniques exist, no technique can achieve perfect diagnosis. As long as a technique might overlook a disease, ophthalmologists must double-check even those images that the technique classifies as normal. Here, we show that our deep-learning-based binary classifier (normal or abnormal) achieved a perfect classification on 108,308 retinal OCT images, i.e., true positive rate = 1.000000 and true negative rate = 1.000000; hence, the area under the ROC curve = 1.0000000 (SOTA performance). Although the test set included three types of diseases, two of these were not used for training. Our work has a sufficient possibility of raising automated diagnosis techniques from “assistant for ophthalmologists” to “independent system without ophthalmologists”.
Keywords: medical image processing, anomaly detection, Artificial intelligence, Retina, OCT, deep learning, Convolutional Neural Network, metric learning, Machine Learning
Suggested Citation: Suggested Citation