Does Machine Unlearning Preserve Clinical Safety? A Risk Analysis for Medical Image Classification

2026-04-26

AI Summary

Existing machine unlearning methods in medical image classification may inadvertently increase false negative rates by neglecting clinical risk, thereby introducing safety hazards. To address this, this work proposes SalUn-CRA, a clinically risk-aware unlearning approach that replaces random relabeling with an entropy-based mechanism to selectively forget malignant samples. This strategy prevents the model from learning spurious benign associations and, for the first time, incorporates asymmetric misdiagnosis costs into the unlearning evaluation framework. Experiments on DermaMNIST and PathMNIST demonstrate that, under 20% and 50% data deletion ratios, SalUn-CRA achieves effective unlearning while maintaining clinical risk significantly lower than or comparable to that of full retraining.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Computer Vision: Adversarial Attacks & Robustness
• Reasoning under Uncertainty: Causality

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

The application of Deep Learning in medical diagnosis must balance patient safety with compliance with data protection regulations. Machine Unlearning enables the selective removal of training data from deployed models. However, most methods are validated primarily through efficiency and privacy-oriented metrics, with limited attention to clinically asymmetric error costs. In this work, we investigate how unlearning affects clinical risk in binary medical image classification. We show that standard unlearning strategies (Fine-Tuning, Random Labeling, and SalUn) may reduce test utility while increasing false-negative rates, thereby amplifying clinical risk. To mitigate this, we propose SalUn-CRA (Clinical Risk-Aware), a variant of SalUn that replaces random relabeling with entropy-based forgetting for malignant samples in the forget set, preventing the model from learning harmful benign associations. We evaluate on DermaMNIST and PathMNIST medical image datasets under 20% and 50% data removal. Using Global Risk metrics with asymmetric costs, SalUn-CRA achieves lower or comparable clinical risk to full retraining while preserving unlearning effectiveness. These results suggest that clinical risk should be an integral component of unlearning validation in medical systems.

Problem

• Machine Unlearning
• Clinical Safety
• Medical Image Classification
• False Negative Rate
• Clinical Risk

Innovation

• Machine Unlearning
• Clinical Risk
• Medical Image Classification
• False-Negative Rate
• Risk-Aware Forgetting