Identification of Homogeneous Subgroups from Resting-State fMRI Data
Hanlu Yang1, Trung Vu1, Qunfang Long1
1Department of Computer Science and Electrical Engineering, University of Maryland Baltimore County, Baltimore, MD 21250, USA.
Sensors (Basel, Switzerland)
|March 30, 2023
Summary
This study introduces a novel framework using functional magnetic resonance imaging (fMRI) data to identify distinct subgroups within psychiatric patients. These subgroups exhibit unique brain activity patterns, aiding in personalized medicine and understanding mental disorders.
Area of Science:
- Neuroscience
- Psychiatry
- Data Science
Background:
- Personalized medicine in psychiatry requires identifying homogeneous patient subgroups.
- Functional connectivity profiles from fMRI are unique but their clinical application in psychiatric disorders is under investigation.
Purpose of the Study:
- To develop and validate a data-driven framework for identifying patient subgroups in psychiatric disorders using fMRI.
- To leverage functional activity maps and the Gershgorin disc theorem for subgroup discovery.
Main Methods:
- A novel pipeline employing constrained independent component analysis based on entropy bound minimization (c-EBM) and eigenspectrum analysis.
- Utilized resting-state network (RSN) templates as constraints for c-EBM to align subject-wise analyses.
- Applied the framework to a large-scale fMRI dataset of 464 psychiatric patients.
Main Results:
- Identified meaningful subgroups within the psychiatric patient cohort.
- Subgroups demonstrated distinct functional connectivity patterns in key brain regions like the dorsolateral prefrontal cortex and anterior cingulate cortex.
- Cognitive test scores significantly differed across identified subgroups, validating their clinical relevance.
Conclusions:
- The proposed framework successfully identifies clinically relevant subgroups in psychiatric disorders using fMRI data.
- This approach advances the characterization of mental disorders and supports the development of personalized treatment strategies.
