Gradient based Severity Labeling for Biomarker Classification in OCT
This addresses the challenge of improving biomarker classification in medical imaging for diseases like Diabetic Retinopathy, but it is incremental as it adapts contrastive learning with a novel selection strategy rather than introducing a new paradigm.
The paper tackles the problem of selecting positive and negative pairs for contrastive learning in medical images, where arbitrary augmentations can distort biomarkers, by proposing a method to generate disease severity labels from gradient responses for unlabeled OCT scans, resulting in up to a 6% improvement in biomarker classification accuracy over self-supervised baselines for Diabetic Retinopathy indicators.
In this paper, we propose a novel selection strategy for contrastive learning for medical images. On natural images, contrastive learning uses augmentations to select positive and negative pairs for the contrastive loss. However, in the medical domain, arbitrary augmentations have the potential to distort small localized regions that contain the biomarkers we are interested in detecting. A more intuitive approach is to select samples with similar disease severity characteristics, since these samples are more likely to have similar structures related to the progression of a disease. To enable this, we introduce a method that generates disease severity labels for unlabeled OCT scans on the basis of gradient responses from an anomaly detection algorithm. These labels are used to train a supervised contrastive learning setup to improve biomarker classification accuracy by as much as 6% above self-supervised baselines for key indicators of Diabetic Retinopathy.