A1438
Title: Streaming tensor decomposition in imaging analysis
Authors: Xiwei Tang - University of Texas at Dallas (United States) [presenting]
Haowen Zhou - University of Virginia (United States)
Abstract: The analysis of complex medical imaging data, particularly from longitudinal or streaming settings, presents significant challenges due to the dynamic and evolving nature of pathological changes. For example, magnetic resonance imaging (MRI) and diffusion imaging (DI) provide critical insights into brain microstructure, but capturing and analyzing the temporal patterns within such multidimensional data streams is a complex task. Tensor decomposition has shown promise in a variety of applications, from neuroscience to social networks, for extracting meaningful patterns. The aim is to present a novel tensor-based framework that combines advanced streaming tensor decomposition techniques with the analysis of longitudinal imaging data. The method is designed to track temporal changes and identify regions that exhibit longitudinal patterns. It is built on tensor tracking, incorporating streaming tensor decompositions to handle the complex, real-time data streams generated by longitudinal imaging. By addressing the unique challenges of temporal structure analysis and data stream factorization, this approach provides new insights into the progression of medical imaging analysis.
