Relational Subsets Knowledge Distillation for Long-tailed Retinal Diseases Recognition

TL;DR

This paper introduces a class subset learning approach with relational knowledge distillation to improve long-tailed retinal disease recognition, effectively handling class imbalance and leveraging prior knowledge for better model performance.

Contribution

It proposes a novel class subset learning framework with relational knowledge distillation tailored for long-tailed medical datasets, enhancing recognition accuracy.

Findings

Significant improvement over baseline models on two datasets.

Effective handling of class imbalance in retinal disease recognition.

Framework is compatible with other state-of-the-art techniques.

Abstract

In the real world, medical datasets often exhibit a long-tailed data distribution (i.e., a few classes occupy most of the data, while most classes have rarely few samples), which results in a challenging imbalance learning scenario. For example, there are estimated more than 40 different kinds of retinal diseases with variable morbidity, however with more than 30+ conditions are very rare from the global patient cohorts, which results in a typical long-tailed learning problem for deep learning-based screening models. In this study, we propose class subset learning by dividing the long-tailed data into multiple class subsets according to prior knowledge, such as regions and phenotype information. It enforces the model to focus on learning the subset-specific knowledge. More specifically, there are some relational classes that reside in the fixed retinal regions, or some common pathological features are observed in both the majority and minority conditions. With those subsets learnt teacher models, then we are able to distill the multiple teacher models into a unified model with weighted knowledge distillation loss. The proposed framework proved to be effective for the long-tailed retinal diseases recognition task. The experimental results on two different datasets demonstrate that our method is flexible and can be easily plugged into many other state-of-the-art techniques with significant improvements.