Generalizable Model to Estimate Individual-Specific Cortical Networks
Poster No:
1632
Submission Type:
Abstract Submission
Authors:
Shreya Pande1, Tianchu Zeng1, Aihuiping Xue2, Ru Kong2, Leon Ooi3, Shaoshi Zhang1, Mervyn Lim1, Christopher Chen1, Juan Helen Zhou1, Phern Chern Tor4, Thomas Yeo5
Institutions:
1National University of Singapore, Singapore, Singapore, 2Computational Brain Imaging Group, Yong Loo Lin School of Medicine, National University of Singapor, Singapore, Singapore, 3National Univeristy of Singapore, Singapore, Singapore, 4National University Hospital, Singapore, Singapore, 5Centre for Sleep and Cognition, National University of Singapore, Singapore, Singapore
First Author:
Co-Author(s):
Aihuiping Xue
Computational Brain Imaging Group, Yong Loo Lin School of Medicine, National University of Singapor
Singapore, Singapore
Computational Brain Imaging Group, Yong Loo Lin School of Medicine, National University of Singapor
Singapore, Singapore
Ruby Kong
Computational Brain Imaging Group, Yong Loo Lin School of Medicine, National University of Singapor
Singapore, Singapore
Computational Brain Imaging Group, Yong Loo Lin School of Medicine, National University of Singapor
Singapore, Singapore
Introduction:
Resting-state functional connectivity (RSFC) identifies large-scale brain networks by measuring synchrony in functional magnetic resonance imaging (fMRI) signals between regions at rest (Biswal et al., 1995). Individual-specific cortical network organization improves behaviour prediction and may serve as a unique fingerprint of human behaviour by capturing properties such as unique size, location and spatial arrangement of brain networks across individuals (Gordon, Laumann, Adeyemo, Gilmore, et al., 2017; Gordon, Laumann, Adeyemo, & Petersen, 2017). Kong et al. (2019) introduced a multi-session hierarchical Bayesian model (MSHBM) to estimate these networks, distinguishing intra-subject (within-subject) and inter-subject (between-subject) network variability. MSHBM parcellations generalize well to new rs-fMRI and task-fMRI data from the same individuals using less data. However, its generalizability to entirely different datasets is limited, requiring retraining of MSHBM for each new dataset.
Methods:
We analysed 14 diverse fMRI datasets (resting-state and task-based) covering a range of demographics, pre-processing methods, and acquisition protocols (Fig. 1a). Various models were tested to generate individual-specific cortical parcellations, with performance evaluated across datasets. The models included:
– Baseline Models: Based on von Mises-Fisher clustering for population-level parcellation(Yeo et al., 2011):
i. Yeo2011_group: Applying group-level parcellations to individual subjects.
ii.Yeo2011_individual: Generating individual-specific parcellations by applying vMF clustering on each individual.
– Baseline MSHBM Models: Using MSHBM for test subjects:
i. HCP_trained: Trained on 40 HCP Young Adults (HCPYA) subjects.
ii. eNKI_trained: Trained on 40 eNKI subjects.
– Multidataset Models: Trained using a leave-one-dataset-out approach, excluding the test dataset:
i. Multidataset_MSHBM: Based on MSHBM 2019.
ii. Multidataset_MSHBM++: An MSHBM 2019 extension with an extra layer to capture dataset-invariant profiles (Fig. 1b).
Each multidataset model was tested with four initialization strategies:
1. No initialization: Derived directly from the training cohort
2. HCPYA initialization: Using group parcellations from HCPYA
3. HCP7T initialization: Using group parcellations from HCP 7T
4. Du Networks: Using group-level Du Networks (Du et al., 2024)
– MSHBM2019 Self Model: Trained on the same dataset as the test set but with subjects excluding the test subjects being used for final evaluation.
– Baseline Models: Based on von Mises-Fisher clustering for population-level parcellation(Yeo et al., 2011):
i. Yeo2011_group: Applying group-level parcellations to individual subjects.
ii.Yeo2011_individual: Generating individual-specific parcellations by applying vMF clustering on each individual.
– Baseline MSHBM Models: Using MSHBM for test subjects:
i. HCP_trained: Trained on 40 HCP Young Adults (HCPYA) subjects.
ii. eNKI_trained: Trained on 40 eNKI subjects.
– Multidataset Models: Trained using a leave-one-dataset-out approach, excluding the test dataset:
i. Multidataset_MSHBM: Based on MSHBM 2019.
ii. Multidataset_MSHBM++: An MSHBM 2019 extension with an extra layer to capture dataset-invariant profiles (Fig. 1b).
Each multidataset model was tested with four initialization strategies:
1. No initialization: Derived directly from the training cohort
2. HCPYA initialization: Using group parcellations from HCPYA
3. HCP7T initialization: Using group parcellations from HCP 7T
4. Du Networks: Using group-level Du Networks (Du et al., 2024)
– MSHBM2019 Self Model: Trained on the same dataset as the test set but with subjects excluding the test subjects being used for final evaluation.
Results:
Models were evaluated using resting-state homogeneity, task-dependent homogeneity, and inhomogeneity (Schaefer et al., 2018). Multidataset models (all initializations) significantly outperformed baseline models(Yeo2011_group, Yeo2011_individual, HCP_trained, and eNKI_trained) and showed comparable performance to the MSHBM2019 self-model, which was trained on the test dataset (Fig. 2). This highlights the generalizability of multidataset-trained models to unseen datasets, in addition to also estimating spatially reliable networks.
Conclusions:
We have proposed an enhanced training approach for MSHBM using a multidataset cohort (varying demographics, acquisition protocols and preprocessing techniques), to achieve a generalizable performance, comparable to dataset-specific models for the test set. Adding layers for dataset-invariant profiles provided limited improvement, as inter-subject variability parameters captured shared features across datasets. Additionally, within-dataset variability exceeded between-dataset variability as evidenced by the concentration parameters used to capture the inter-dataset and inter-subject variability. Extending MSHBM 2019 to generalize across diverse datasets improves adaptability, addressing demographic, acquisition, and pre-processing differences. This robust framework enhances cortical network modelling, advancing neuroscience research and clinical applications to more diverse datasets.
Modeling and Analysis Methods:
Bayesian Modeling
Connectivity (eg. functional, effective, structural)
fMRI Connectivity and Network Modeling 2
Methods Development
Segmentation and Parcellation 1
Keywords:
Computational Neuroscience
Cortex
FUNCTIONAL MRI
Modeling
Other - Functional Network Estimation; Resting-state homogeneity; Task-dependent homogeneity; Task-dependent inhomogeneity; Parcellation; Generalizability
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?
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, J., DiNicola, L. M., Angeli, P. A., Saadon-Grosman, N., Sun, W., Kaiser, S., Ladopoulou, J., Xue, A., Yeo, B. T. T., Eldaief, M. C., & Buckner, R. L. (2024). Organization of the human cerebral cortex estimated within individuals: Networks, global topography, and function. Journal of Neurophysiology.
Gordon, E. M., Laumann, T. O., Adeyemo, B., Gilmore, A. W., Nelson, S. M., Dosenbach, N. U. F., & Petersen, S. E. (2017). Individual-specific features of brain systems identified with resting state functional correlations. NeuroImage, 146, 918–939.
Gordon, E. M., Laumann, T. O., Adeyemo, B., & Petersen, S. E. (2017). Individual Variability of the System-Level Organization of the Human Brain. Cerebral Cortex (New York, N.Y.: 1991), 27(1), 386–399.
Kong, R., Li, J., Orban, C., Sabuncu, M. R., Liu, H., Schaefer, A., Sun, N., Zuo, X.-N., Holmes, A. J., Eickhoff, S. B., & Yeo, B. T. T. (2019). Spatial Topography of Individual-Specific Cortical Networks Predicts Human Cognition, Personality, and Emotion. Cerebral Cortex (New York, N.Y.: 1991), 29(6), 2533–2551.
Schaefer, A., Kong, R., Gordon, E. M., Laumann, T. O., Zuo, X.-N., Holmes, A. J., Eickhoff, S. B., & Yeo, B. T. T. (2018). Local-Global Parcellation of the Human Cerebral Cortex from Intrinsic Functional Connectivity MRI. Cerebral Cortex (New York, NY), 28(9), 3095–3114.
Yeo, B. T. T., Krienen, F. M., Sepulcre, J., Sabuncu, M. R., Lashkari, D., Hollinshead, M., Roffman, J. L., Smoller, J. W., Zöllei, L., Polimeni, J. R., Fischl, B., Liu, H., & Buckner, R. L. (2011). The organization of the human cerebral cortex estimated by intrinsic functional connectivity. Journal of Neurophysiology, 106(3), 1125–1165.