Abstract
Group Independent Component Analysis (gICA) is widely used in clinical neu-roimaging to decompose high-dimensional functional MRI data into interpretable brain networks. However, conventional gICA primarily models components that are shared across subjects. This group-level assumption can limit the recovery of networks present only in individual subjects or subsets of subjects, reducing sen-sitivity to inter-subject heterogeneity in clinical and heterogeneous neuroimaging datasets. We introduce Copula-Linked Group ICA (CoLiG-ICA), a novel algorithm within the Group ICA 2.0 framework that jointly estimates template-linked, co-hort-only, and subject-only brain networks within a unified model. The proposed algorithm, CoLiG-ICA, combines ICA-based spatial decomposition, copula-based dependence modeling, and deep learning optimization to retain the con-sistency and interpretability of template-constrained ICA while enabling addition-al free components beyond the provided reference networks. By linking subject decompositions to shared templates and jointly estimating cohort-only and sub-ject-only sources, CoLiG-ICA provides a flexible representation of individual variability that is not captured by conventional group priors. We evaluate CoLiG-ICA using resting-state fMRI data from the UCLA Con-sortium for Neuropsychiatric Phenomics dataset. We compare the proposed algo-rithm with conventional constrained ICA to assess estimation of template-linked components, discovery of additional free components, component independence, and the degree to which the estimated components capture subject-level variability beyond the shared group prior. In a schizophrenia-only group analysis, CoLiG-ICA identified three additional resting-state components beyond the 53 template-linked NeuroMark components. Compared with MOO-ICAR, CoLiG-ICA also showed significantly lower inter-component spatial dependence, indicating improved subject-level component in-dependence, and significantly reduced motion-related variance in the template-linked components.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/BrainWorks_010.pdf
SharedIt Link: Not yet available
SpringerLink (DOI): Not yet available
Supplementary Material: Not Submitted
Link to Open Review
Open Review Page: https://openreview.net/profile?id=%7EOktay_Agcaoglu1
BibTex
@InProceedings{AgcOkt_Group_MICCAISAT2026,
author = { Agcaoglu, Oktay},
title = { { Group ICA 2.0: Closing the Gap Between Subjects and Group Latent Decomposition } },
booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026 Workshops and Challenges},
year = {2026},
publisher = {Springer Nature Switzerland},
volume = {LNCS 17254},
month = {pending},
page = {pending}
}