GRIN+ enables fast effective data removal in imbalanced medical AI
GRIN+: Towards Fast Yet Effective Machine Unlearning for Imbalanced Medical Data
Artificial Intelligence
Summary
When hospitals use AI to help diagnose patients, sometimes sensitive data must be removed from these AI systems to protect privacy laws like GDPR. Existing methods to do this can be slow or accidentally harm the AI’s ability to detect rare diseases because they overlook important but uncommon data. The authors present GRIN+, a new technique that quickly and carefully removes private patient data while keeping the AI accurate on all medical cases, especially those that happen rarely. They tested it on datasets like skin cancer and brain tumors, showing it balances privacy, speed, and medical accuracy better than previous methods.
What this means in practice
- •For hospital data teams: Remove specific patient data from medical AI models quickly without harming accuracy on rare diseases.
- •For healthcare ai developers: Build privacy-compliant medical AI systems that maintain performance despite class imbalances in training data.
Authors
Minghui Huang, Junxiao Wang
Abstract
As deep learning models become fundamental to modern healthcare, the "Right to be Forgotten" mandated by privacy regulations like GDPR and HIPAA necessitates effective machine unlearning (MU) to remove sensitive patient data from trained models. However, existing MU techniques often struggle with a fundamental "privacy-efficiency-utility" (PEU) trilemma, particularly in medical scenarios where data is frequently characterized by severe class imbalance and long-tailed distributions. In such cases, standard unlearning methods can fail to protect key clinical knowledge or mistakenly delete features essential for diagnosing rare conditions due to the gradient dominance of majority classes. To address these challenges, we propose GRIN+, a novel machine unlearning framework designed for fast and precise data erasure in imbalanced medical scenarios. GRIN+ decouples unlearning-specific knowledge from generalized representations at the parameter level by analyzing the gradient contributions of both "forget" and "retain" sets. It introduces a class-adaptive influence scoring mechanism to rectify gradient dominance and employs a direction-constrained update strategy to prevent the unintended erosion of vital clinical knowledge. Comprehensive benchmarking across multiple medical datasets, including skin cancer (ISIC), brain tumor (MRI), and breast ultrasound (BUSI), demonstrates that GRIN+ achieves an optimal balance of the PEU trilemma. Experimental results show that GRIN+ maintains high diagnostic accuracy and robust privacy while significantly enhancing runtime efficiency compared to existing baselines. We open-source the GRIN+ code and benchmarks to support further research.