Extraction and Automatic Grouping of Joint and Individual Sources in Multisubject fMRI Data Using Higher Order...

Area of Science:

• Biomedical and Health Informatics
• Neuroimaging Analysis
• Multivariate Data Analysis

Background:

• Analyzing interdependencies across multiple datasets is crucial in biomedical and health informatics.
• Existing methods may struggle to effectively distinguish between joint and individual sources in multisubject data.

Purpose of the Study:

• To propose a novel algorithm, joint/individual thin independent component analysis (JI-ThICA), for extracting joint and individual sources from multisubject data.
• To develop a method that automatically determines the number and types of sources and categorizes similar sources across datasets.

Main Methods:

• JI-ThICA employs a deflation-based procedure utilizing higher-order cumulants for source extraction.
• Joint sources are identified by fusing signals from all subjects, while individual sources are extracted separately per subject.
• The algorithm incorporates an efficient strategy for estimating the number of joint sources and an optimization process for source categorization.

Main Results:

• Simulated fMRI data analysis showed JI-ThICA outperforms existing alternatives in mean joint signal to interference ratio, even with noisy signals.
• Application to a real-world fMRI dataset from a naturalistic auditory experience yielded results consistent with prior studies.
• The algorithm successfully extracted reliable and meaningful information from multisubject fMRI data.

Conclusions:

• JI-ThICA is an effective algorithm for disentangling joint and individual sources in multisubject data.
• The method demonstrates robust performance in analyzing both simulated and real fMRI datasets.
• JI-ThICA offers a valuable tool for advancing research in biomedical and health informatics, particularly in neuroimaging.