Search papers, labs, and topics across Lattice.
The paper introduces a gradient-based severity labeling method for Optical Coherence Tomography (OCT) scans to improve contrastive learning for biomarker classification. They generate pseudo-labels representing disease severity using gradient responses from an anomaly detection algorithm applied to unlabeled OCT images. These pseudo-labels are then used to guide the selection of positive and negative pairs in a supervised contrastive learning framework, leading to improved biomarker classification performance.
Severity-based contrastive learning, guided by gradients from anomaly detection, boosts biomarker classification in OCT scans where traditional augmentation-based methods falter.
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.
