Multi-center MRI Image Feature Alignment and Alzheimer's Disease Screening Based on Transfer Learning

Abstract

Alzheimer's disease, as a major public health challenge faced by the global aging society, its early screening and precise diagnosis are of crucial significance for delaying the progression of the disease and reducing medical costs. Magnetic resonance imaging has become a core tool for diagnosing Alzheimer's disease because it can clearly present changes in brain structure. However, problems such as distribution differences and feature heterogeneity in multi-center data make it difficult for traditional models to be directly applied across centers. Transfer learning provides a new approach to solving the problem of data imbalance in multi-center MRI image analysis by extracting cross-domain shared features. This paper systematically reviews the application framework of transfer learning in multi-center MRI feature alignment, analyzes its technical advantages and practical value in Alzheimer's disease screening, and explores the promoting effect of cross-center collaboration models on improving diagnostic efficiency. Research shows that transfer learning can effectively narrow the distribution differences among multi-center data and significantly improve the accuracy and generalization ability of early screening for Alzheimer's disease.