Masked Image Modelling for retinal OCT understanding

TL;DR

This paper demonstrates that masked autoencoders can effectively learn representations of retinal OCT images, improving performance on multiple tasks and enabling multimodal fusion with IR fundus images, using a large-scale dataset.

Contribution

It introduces the first extensive evaluation of masked image modelling for OCT, and extends MAE pretraining to multimodal fusion with IR fundus images for improved performance.

Findings

Strong performance on 6 downstream tasks after fine-tuning

Effective as a frozen feature extractor with lightweight adapters

Improved multimodal performance with joint OCT and IR model

Abstract

This work explores the effectiveness of masked image modelling for learning representations of retinal OCT images. To this end, we leverage Masked Autoencoders (MAE), a simple and scalable method for self-supervised learning, to obtain a powerful and general representation for OCT images by training on 700K OCT images from 41K patients collected under real world clinical settings. We also provide the first extensive evaluation for a model of OCT on a challenging battery of 6 downstream tasks. Our model achieves strong performance when fully finetuned but can also serve as a versatile frozen feature extractor for many tasks using lightweight adapters. Furthermore, we propose an extension of the MAE pretraining to fuse OCT with an auxiliary modality, namely, IR fundus images and learn a joint model for both. We demonstrate our approach improves performance on a multimodal downstream application. Our experiments utilize most publicly available OCT datasets, thus enabling future comparisons. Our code and model weights are publicly available https://github.com/TheoPis/MIM_OCT.