Common Subspace Identification

Functional magnetic resonance imaging (fMRI) has been widely used to study functional connectivity of the brain. The functional connectivity profiles captured by fMRI preserve subject variability and hence act as fingerprints. This motivates the use of functional connectivity networks and their relationship to be used in the identification of subgroups of subjects, e.g., corresponding to subtypes of disease or mental disorders. This is also an effective way to summarize the heterogeneity in large datasets.

The provided package presents a method based on independent vector analysis (IVA) to identify subgroups from muti-subject fMRI data. The method leverages the strength of IVA to identify structures in multiset data with strong identifiability guarantees. Subgroup structures are identified based on spatial functional connectivity across subjects. Additionally, a novel subgroup identification method, utilizing Gershgorin disc, is included to deliver a parameter-free solution for identifying homogenous subgroups.

The original methodology is introduced in [1] and Gershgorin disc [2] has been successfully employed in research related to subgroup analysis [2,3,4], yielding promising and insightful outcomes. The package automatically detects subgroups of subjects from IVA results in three steps:
1. Classification of source component vectors (SCVs) into three groups:
  • Common SCVs with highly correlated components
  • Group-specific SCVs with subgroup structure
  • Distinct SCVs with no discerniblestructure (low correlation values)
2. Clustering of SCVs within the second group into N_cluster_SCV clusters.
3. Identification of subject subgroups for each cluster. Three different subgroup identification methods are implemented:
  • Modularization
  • Clustering
  • Gershgorin disc
    Independent Vector Analysis for Common Subspace Extraction (IVA-CS) Matlab Package


References:

[1] Long, Qunfang, et al., "Independent vector analysis for common subspace analysis: Application to multi-subject fMRI data yields meaningful subgroups of schizophrenia," NeuroImage 216 (2020): 116872.
[2] Yang, Hanlu, et al., "Independent vector analysis based subgroup identification from multisubject fMRI data," ICASSP, IEEE, 2022.
[3] Yang, Hanlu, et al., "Constrained Independent Component Analysis Based on Entropy Bound Minimization for Subgroup Identification from Multi-subject fMRI Data," ICASSP, IEEE, 2023.
[4] Yang, Hanlu, et al., "Identification of Homogeneous Subgroups from Resting-State fMRI Data," Sensors 23.6 (2023): 3264.