Test-Time Selection for Robust Skin Lesion Analysis

TL;DR

This paper introduces TTS, a human-in-the-loop test-time selection method that reduces bias from artifacts in skin lesion analysis models without retraining, using minimal annotations and effectively handling varying bias levels.

Contribution

The paper presents TTS, a novel test-time selection approach that leverages keypoints to mitigate artifact bias in skin lesion models without retraining or extensive annotations.

Findings

TTS effectively reduces artifact bias in skin lesion analysis.

TTS performs well with limited and varying annotations.

Application on ISIC2019 demonstrates real-world deployment potential.

Abstract

Skin lesion analysis models are biased by artifacts placed during image acquisition, which influence model predictions despite carrying no clinical information. Solutions that address this problem by regularizing models to prevent learning those spurious features achieve only partial success, and existing test-time debiasing techniques are inappropriate for skin lesion analysis due to either making unrealistic assumptions on the distribution of test data or requiring laborious annotation from medical practitioners. We propose TTS (Test-Time Selection), a human-in-the-loop method that leverages positive (e.g., lesion area) and negative (e.g., artifacts) keypoints in test samples. TTS effectively steers models away from exploiting spurious artifact-related correlations without retraining, and with less annotation requirements. Our solution is robust to a varying availability of annotations, and different levels of bias. We showcase on the ISIC2019 dataset (for which we release a subset of annotated images) how our model could be deployed in the real-world for mitigating bias.