Analise de Desaprendizado de Maquina em Modelos de Classificacao de Imagens Medicas

2025-08-25

AI Summary

This work addresses the critical need for privacy-preserving removal of sensitive data from medical image classification models. We introduce, for the first time, machine unlearning into medical imaging, proposing an efficient unlearning method based on the SalUn algorithm. We systematically evaluate its efficacy on three public medical image datasets—PathMNIST, OrganAMNIST, and BloodMNIST—and investigate how data augmentation enhances unlearning quality and model robustness. Experiments demonstrate that SalUn achieves unlearning performance comparable to full retraining while requiring only local model updates, thereby significantly reducing computational overhead without compromising classification accuracy or generalization capability. This study bridges a key gap in applying machine unlearning to medical imaging and establishes a reproducible, scalable technical framework for governing healthcare AI models under stringent privacy requirements.

Technology Category

• Machine Learning: Privacy
• Computer Vision: Adversarial Attacks & Robustness
• Humans and AI: Human-in-the-loop Machine Learning

Application Category

• Security and Privacy: Security and privacy of machine learning and AI applications
• User Modeling, Personalization and Recommendation: User privacy protection in personalized systems
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Machine unlearning aims to remove private or sensitive data from a pre-trained model while preserving the model's robustness. Despite recent advances, this technique has not been explored in medical image classification. This work evaluates the SalUn unlearning model by conducting experiments on the PathMNIST, OrganAMNIST, and BloodMNIST datasets. We also analyse the impact of data augmentation on the quality of unlearning. Results show that SalUn achieves performance close to full retraining, indicating an efficient solution for use in medical applications.

Problem

• Evaluating machine unlearning for medical image classification
• Assessing data augmentation impact on unlearning quality
• Validating SalUn performance against full retraining benchmarks

Innovation

• SalUn unlearning model for medical images
• Evaluated on multiple medical datasets
• Achieves performance close to full retraining