Copula Linked Group ICA: Capturing variable coupling between subject and group brain networks
Poster No:
1481
Submission Type:
Abstract Submission
Authors:
Oktay Agcaoglu1, Rogers Silva2, Vince Calhoun3
Institutions:
1Tri-Institutional Center for Translational Research in Neuroimaging and Data Science (TReNDS), Atlanta, GA, 2TReNDS Center, Atlanta, GA, 3GSU/GATech/Emory, Atlanta, GA
First Author:
Oktay Agcaoglu, PhD
Tri-Institutional Center for Translational Research in Neuroimaging and Data Science (TReNDS)
Atlanta, GA
Tri-Institutional Center for Translational Research in Neuroimaging and Data Science (TReNDS)
Atlanta, GA
Co-Author(s):
Introduction:
fMRI is a powerful tool for investigating brain connectivity but is inherently high-dimensional and noisy. Group independent component analysis (gICA) is commonly used to reduce dimensionality while accounting for subject variability and removing noise such as motion. Traditional gICA decomposes the data into components common across all subjects and does not explicitly model components that are specific to subjects, potentially impacting its ability to fully capture subject variability. We propose a novel gICA algorithm that estimates subject-specific components while simultaneously linking components to aggregated group components or priors while allowing for varying degrees of similarity. This approach also allows for the inclusion of unlinked, subject-specific components, effectively capturing greater subject variability and enhancing the precision of connectivity analyses.
Methods:
Conventional ICA algorithms, such as Infomax and FastICA, are based on predefined learning rules that limit their flexibility in modeling complex cost functions. These limitations make it challenging to incorporate advanced loss functions that capture intricate relationships between modalities or prior information. Recent advances in computational methods, particularly deep learning frameworks (DL-F) with automatic differentiation, have opened the door to optimizing more sophisticated loss functions. Building on these advancements, we recently introduced a copula framework for neuroimaging data (Agcaoglu,2024), which combines the strengths of copulas, ICA, and DL-F. Expanding on this, we propose a novel gICA algorithm, named Copula-Linked gICA (CoLiG), which enables both group-level and subject-specific independent components. We show the approach has several advantages over existing approaches. To demonstrate, we used CoLiG to incorporate spatial priors, namely the NeuroMark ICA (NM) brain networks (Du,2020), while estimating both subject common and subject specific components in a large fMRI dataset of schizophrenia patients (SZ) and controls (CN) data (Keator,2016). The NM templates include 53 resting-state networks, and we selected a model order of 100, allowing for 47 subject-specific components. Each subject's preprocessed fMRI data was reduced to voxel-by-100 dimensions with PCA. 100x100 unmixing matrix was then estimated, with 53 of these components linked to NM templates using a Gaussian copula, while the remaining 47 components were left unlinked. We analyzed data from 311 subjects, including 160 CT and 151 SZ individuals. Gaussian copula dependency values are set to 0.98 for all components.
Results:
Preliminary results demonstrated that the proposed CoLiG method captured greater subject variability, resulting in higher connectivity across nearly all domains, particularly in subcortical (SB), sensorimotor (SM), visual (VIS), and cerebellar (CB) regions. Additionally, CoLiG-ICA detected more group differences between CN and SZ groups, as assessed by two-sample t-tests with false discovery rate correction (p < 0.01). When compared with the MOO-ICAR algorithm, CoLiG identified a greater number of group differences, further highlighting its effectiveness. Notably, meaningful components were also identified among the unlinked components, extending beyond predefined group networks or priors and emphasizing the robustness of the CoLiG method.
Conclusions:
The proposed CoLiG-ICA algorithm advances traditional gICA by capturing both group-level and subject-specific components, significantly enhancing the precision of connectivity analyses. By incorporating spatial priors while allowing unlinked components, CoLiG effectively captures greater subject variability and provides insights beyond predefined group networks. Preliminary results also revealed meaningful subject-specific components, further emphasizing the method's ability to capture unique individual variability. These findings demonstrate CoLiG's potential as a powerful tool for neuroimaging studies.
Disorders of the Nervous System:
Psychiatric (eg. Depression, Anxiety, Schizophrenia) 2
Modeling and Analysis Methods:
Bayesian Modeling
Connectivity (eg. functional, effective, structural)
fMRI Connectivity and Network Modeling 1
Task-Independent and Resting-State Analysis
Keywords:
FUNCTIONAL MRI
Machine Learning
Modeling
Schizophrenia
1|2Indicates the priority used for review
Abstract Information
By submitting your proposal, you grant permission for the Organization for Human Brain Mapping (OHBM) to distribute your work in any format, including video, audio print and electronic text through OHBM OnDemand, social media channels, the OHBM website, or other electronic publications and media.
The Open Science Special Interest Group (OSSIG) is introducing a reproducibility challenge for OHBM 2025. This new initiative aims to enhance the reproducibility of scientific results and foster collaborations between labs. Teams will consist of a “source” party and a “reproducing” party, and will be evaluated on the success of their replication, the openness of the source work, and additional deliverables. Click here for more information. Propose your OHBM abstract(s) as source work for future OHBM meetings by selecting one of the following options:
Please indicate below if your study was a "resting state" or "task-activation” study.
Healthy subjects only or patients (note that patient studies may also involve healthy subjects):
Was this research conducted in the United States?
Are you Internal Review Board (IRB) certified? Please note: Failure to have IRB, if applicable will lead to automatic rejection of abstract.
Were any human subjects research approved by the relevant Institutional Review Board or ethics panel? NOTE: Any human subjects studies without IRB approval will be automatically rejected.
Were any animal research approved by the relevant IACUC or other animal research panel? NOTE: Any animal studies without IACUC approval will be automatically rejected.
Please indicate which methods were used in your research:
For human MRI, what field strength scanner do you use?
Which processing packages did you use for your study?
Provide references using APA citation style.
Du, Y., Fu, Z., Sui, J., Gao, S., Xing, Y., Lin, D., Salman, M., Abrol, A., Rahaman, M. A., Chen, J., Hong, L. E., Kochunov, P., Osuch, E. A., Calhoun, V. D., & Alzheimer's Disease Neuroimaging, I. (2020). NeuroMark: An automated and adaptive ICA based pipeline to identify reproducible fMRI markers of brain disorders. Neuroimage Clin, 28, 102375. https://doi.org/10.1016/j.nicl.2020.102375
Keator, D. B., van Erp, T. G. M., Turner, J. A., Glover, G. H., Mueller, B. A., Liu, T. T., Voyvodic, J. T., Rasmussen, J., Calhoun, V. D., Lee, H. J., Toga, A. W., McEwen, S., Ford, J. M., Mathalon, D. H., Diaz, M., O'Leary, D. S., Jeremy Bockholt, H., Gadde, S., Preda, A., . . . Fbirn. (2016). The Function Biomedical Informatics Research Network Data Repository. Neuroimage, 124(Pt B), 1074-1079. https://doi.org/10.1016/j.neuroimage.2015.09.003