Expert-Guided Explainable Few-Shot Learning with Active Sample Selection for Medical Image Analysis

TL;DR

This paper introduces a dual-framework combining expert-guided explainable few-shot learning and explainability-guided active learning to improve medical image analysis, achieving high accuracy with limited data and enhanced interpretability.

Contribution

It proposes a novel integrated approach that leverages radiologist expertise and explainability to guide data selection and model training in medical imaging tasks.

Findings

Achieves over 90% accuracy on MRI data with few samples.

Outperforms non-guided baselines across multiple datasets.

Demonstrates improved interpretability with Grad-CAM visualizations.

Abstract

Medical image analysis faces two critical challenges: scarcity of labeled data and lack of model interpretability, both hindering clinical AI deployment. Few-shot learning (FSL) addresses data limitations but lacks transparency in predictions. Active learning (AL) methods optimize data acquisition but overlook interpretability of acquired samples. We propose a dual-framework solution: Expert-Guided Explainable Few-Shot Learning (EGxFSL) and Explainability-Guided AL (xGAL). EGxFSL integrates radiologist-defined regions-of-interest as spatial supervision via Grad-CAM-based Dice loss, jointly optimized with prototypical classification for interpretable few-shot learning. xGAL introduces iterative sample acquisition prioritizing both predictive uncertainty and attention misalignment, creating a closed-loop framework where explainability guides training and sample selection synergistically. On the BraTS (MRI), VinDr-CXR (chest X-ray), and SIIM-COVID-19 (chest X-ray) datasets, we achieve accuracies of 92\%, 76\%, and 62\%, respectively, consistently outperforming non-guided baselines across all datasets. Under severe data constraints, xGAL achieves 76\% accuracy with only 680 samples versus 57\% for random sampling. Grad-CAM visualizations demonstrate guided models focus on diagnostically relevant regions, with generalization validated on breast ultrasound confirming cross-modality applicability.