Self-supervised contrastive learning improves machine learning discrimination of full thickness macular holes from epiretinal membranes in retinal OCT scans

Abstract

There is a growing interest in using computer-assisted models for the detection of macular conditions using optical coherence tomography (OCT) data. As the quantity of clinical scan data of specific conditions is limited, these models are typically developed by fine-tuning a generalized network to classify specific macular conditions of interest. Full thickness macular holes (FTMH) present a condition requiring timely surgical intervention to prevent permanent vision loss. Other works on automated FTMH classification have tended to use supervised ImageNet pre-trained networks with good results but leave room for improvement. In this paper, we develop a model for FTMH classification using OCT slices around the central foveal region to pre-train a naïve network using contrastive self-supervised learning. We found that self-supervised pre-trained networks outperform ImageNet pre-trained networks despite a small training set size (284 eyes total, 51 FTMH+ eyes, 3 slices from each eye). 3D spatial contrast pre-training yields a model with an F1- score of 1.0 on holdout data (50 eyes total, 10 FTMH+), compared ImageNet pre-trained models, respectively. These results demonstrate that even limited data may be applied toward self-supervised pre- training to substantially improve performance for FTMH classification, indicating applicability toward other OCT-based problems.