Uncertainty-aware deep learning methods for robust diabetic retinopathy classification

TL;DR

This paper investigates uncertainty estimation methods in deep learning for diabetic retinopathy classification, extending analysis from binary to multi-class schemes and proposing a new uncertainty measure with clinical relevance.

Contribution

It introduces a systematic evaluation of uncertainty measures on clinical datasets and develops a novel uncertainty metric linked to classifier risk, applicable to multi-class classification.

Findings

Entropy-based uncertainty generalizes to binary but not multi-class classification.

The new uncertainty measure generalizes effectively to 5-class classification.

Systematic analysis on clinical datasets enhances understanding of uncertainty in medical AI.

Abstract

Automatic classification of diabetic retinopathy from retinal images has been widely studied using deep neural networks with impressive results. However, there is a clinical need for estimation of the uncertainty in the classifications, a shortcoming of modern neural networks. Recently, approximate Bayesian deep learning methods have been proposed for the task but the studies have only considered the binary referable/non-referable diabetic retinopathy classification applied to benchmark datasets. We present novel results by systematically investigating a clinical dataset and a clinically relevant 5-class classification scheme, in addition to benchmark datasets and the binary classification scheme. Moreover, we derive a connection between uncertainty measures and classifier risk, from which we develop a new uncertainty measure. We observe that the previously proposed entropy-based uncertainty measure generalizes to the clinical dataset on the binary classification scheme but not on the 5-class scheme, whereas our new uncertainty measure generalizes to the latter case.