🤖 AI Summary
This study presents the first systematic evaluation of instance-level machine unlearning methods for 3D medical image segmentation, addressing the need to efficiently remove data from specific individuals in compliance with privacy regulations such as GDPR. Leveraging a Med3D-pretrained 3D ResNet-50 backbone, four approximate unlearning strategies are compared on the MRBrainS18 dataset, with performance trade-offs between forgetting efficacy and retained utility quantified using Dice coefficient and mean absolute error (MAE). Experimental results demonstrate that the Noisy Label strategy reduces performance on the forget set by 93% after 50 training epochs while maintaining 84% accuracy on the retain set—significantly outperforming other approaches, which all suffer catastrophic degradation in retained performance under prolonged training. This work establishes a rigorous baseline for the task and identifies an optimal balance between effective unlearning and model utility.
📝 Abstract
With new data privacy laws such as the General Data Protection Regulation (GDPR) [1] that allow individuals to ask that any of their personal information be erased from trained machine learning models, there has been a push to investigate the unlearning of data from models as a way to comply with these laws. In this regard, based on four mechanics, we consider several approximate unlearning strategies applied to the MRBrainS18 dataset [2]. We use a 3D ResNet-50 [3] as a backbone architecture for segmentation that has been pre-trained with the Med3D framework [4]. Considering the pre-trained model as a baseline, we evaluate respective retention accuracy on 2 types of subjects, i.e., retain and forget. We assess these approaches through their Dice similarity coefficient and mean absolute error (MAE) values using two separate training horizons 20 and 50 epochs. The results show that the Noisy Label strategy had the best overall trade-off with a decrease of 93% in the forget set while maintaining 84% accuracy for the retained set after 50 epochs. All other strategies showed extreme levels of forgetting at higher epoch numbers while also demonstrating catastrophic degradation of their retain set performance. The results of this study provide a strict baseline of performance metrics for unlearning on a subject-specific level and provide practitioners with clear criteria for selecting the proper strategies.