Brain Age Estimation Using Structural MRI: A Clustered Federated Learning Approach.
COINS(2023)
摘要
Estimating brain age based on structural Magnetic Resonance Imaging (MRI) is one of the most challenging and prominent areas of research in recent medical imaging and neuroscience studies. The significance of brain age prediction in the early diagnosis of neurological disorders has fueled a resurgence of interest in this field. Various studies have addressed this issue using a spectrum of techniques, from traditional machine learning to deep neural networks. The majority of these techniques employ centralized paradigms, which do not adequately preserve privacy. To tackle this problem, a handful of studies have utilized a federated approach. In this study, we propose a novel hierarchical clustered federated learning approach that carefully captures and considers the similarities of the clients’ predictions on a certain benchmark dataset. This method enhances performance in Non-Independent and Identically Distributed (Non-IID) environments while preserving privacy. We use a multi-site dataset that provides a broad variety of MRI scans, characterizing a proper Non-IID environment. Our method achieves a Mean Absolute Error (MAE) of 3.86, while the non-clustered FedAvg federated approach attains a 4.14 MAE on the test set.
1
1
Code is available at https://github.com/Abtinmy/Clustered-FL-BrainAGE.
更多查看译文
关键词
Brain Age Estimation,Privacy-Preserving Machine Learning (PPML),Federated Learning (FL),Clustering
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络
Chat Paper
正在生成论文摘要