Atlas-Free Functional Brain Connectome Analysis via Task-Driven Parcellation
Poster No:
1482
Submission Type:
Late-Breaking Abstract Submission
Authors:
Keqi Han1, Yao Su2, Songlin Zhao3, Charles Gillespie1, Boadie Dunlop1, Daniel Barron4, Randy Hirschtick5, Liang Zhan6, Lifang He3, Xiang Li5, Carl Yang1
Institutions:
1Emory University, Atlanta, GA, 2Worcester Polytechnic Institute, Worcester, MA, 3Lehigh University, Bethlehem, PA, 4Brigham and Women's Hospital, Boston, MA, 5Massachusetts General Hospital, Boston, MA, 6University of Pittsburgh, Pittsburgh, PA
First Author:
Co-Author(s):
Late Breaking Reviewer(s):
Introduction:
Selecting a suitable brain atlas for node definition is a critical yet challenging step in functional connectome analysis (Stanley, 2013). A mismatched atlas can obscure subtle topographies and undermine the subsequent analysis. In this work, we propose AFCON, an atlas-free method that bypasses atlas selection by jointly optimizing an adaptive parcellation module and a graph-based connectome analysis module. AFCON adaptively generates task-specific, individualized parcellations from fMRI data, which better aligns with the prediction task and offers enhanced interpretability. We also introduce two neurobiologically-informed regularizers to ensure plausible parcellations. Experiments on ADHD (ADHD-200 consortium, 2012) and ADNI (Mueller, 2005) show that AFCON outperforms standard atlas-based baselines in prediction, and reveals regional biomarkers that are consistent with their established roles in neural pathology. Notably, this work focuses on the cerebral cortex, serving as an initial step towards potential whole brain connectivity analysis in the future for more robust clinical utility.
Methods:
AFCON integrates a 3D U-Net (Ronneberger, 2015) for adaptive brain parcellation with a graph neural network for clinical prediction. The 3D U-Net segments cortical voxels into K non-overlapping ROIs, refined by two regularizers. The balanced distribution regularizer computes the KL-divergence between the ROI assignment distribution and a uniform distribution to ensure comparable parcel sizes, while the spatial compactness regularizer minimizes the variance of voxel coordinates within each ROI to encourage contiguous, anatomically consistent regions. ROI-level signals are averaged to construct a functional connectome that is fed into the GCN (Kipf, 2016) for clinical outcome prediction. We evaluate AFCON on two rs-fMRI datasets: ADHD-200 (569 subjects in total, with 246 ADHD patients) and ADNI (200 subjects in total, with 100 AD patients), both datasets are processed with standard fMRIPrep pipeline (Esteban, 2019), including skull stripping, spatial normalization, segmentation, slice-time correction, susceptibility distortion correction, and motion artifact removal. We compare AFCON against atlas-based brain connectome analysis baselines with four well-established atlases: Harvard-Oxford48 (Makris, 2006), AAL90 (Tzourio-Mazoyer, 2002), Schaefer200 (Schaefer, 2018) and HCP360 (Glasser, 2016).
Results:
For the disease prediction task, AFCON outperforms atlas-based baselines across multiple parcellation granularities (#ROIs=48, 90, 200, 360) on both the ADHD and ADNI datasets. Quantitative evaluation of the learned parcellations shows that AFCON achieves higher region homogeneity than both well-established atlases and random parcellations, indicating more functionally cohesive ROIs. Additionally, AFCON's parcellations achieve strong inter-subject consistency, preserving shared brain structures across subjects while accommodating individual variability, ensuring robustness for group-level analysis. Qualitative analysis validates AFCON's effectiveness in biomarker identification, revealing that functionally significant ROIs (p<0.05), such as the Precentral, Occipital, Fusiform, and Cingulum for ADHD, and the Precuneus, Inferior Parietal, and Middle Frontal regions for AD, which are consistent with existing neuroscientific findings. An ablation study validates the regularizers' role, showing that their removal degrades performance and disrupts plausibility of parcellations.
Conclusions:
In this study, we introduce AFCON, an atlas-free framework for functional brain connectome analysis that integrates adaptive parcellation with downstream prediction. By jointly optimizing parcellation and prediction, AFCON outperforms classical atlas-based methods and reveals disease-relevant biomarkers in ADHD and AD cohorts. The proposed framework demonstrates the potential to advance clinical diagnosis and treatment by identifying task-specific brain signatures without predefined atlases.
Modeling and Analysis Methods:
Classification and Predictive Modeling
Connectivity (eg. functional, effective, structural)
fMRI Connectivity and Network Modeling 1
Methods Development
Segmentation and Parcellation 2
Keywords:
Computational Neuroscience
Design and Analysis
FUNCTIONAL MRI
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:
Provide references using APA citation style.
[2] Esteban, O. (2019). fMRIPrep: a robust preprocessing pipeline for functional MRI. Nature methods, 16(1), 111-116.
[3] Glasser, M. F. (2016). A multi-modal parcellation of human cerebral cortex. Nature, 536(7615), 171-178.
[4] Kipf, T. N. (2016). Semi-supervised classification with graph convolutional networks. arXiv preprint arXiv:1609.02907.
[5] Makris, N. (2006). Decreased volume of left and total anterior insular lobule in schizophrenia. Schizophrenia research, 83(2-3), 155-171.
[6] Mueller, S. G. (2005). The Alzheimer's disease neuroimaging initiative. Neuroimaging Clinics, 15(4), 869-877.
[7] Ronneberger, O. (2015). U-net: Convolutional networks for biomedical image segmentation. In Medical image computing and computer-assisted intervention–MICCAI 2015: 18th international conference, Munich, Germany, October 5-9, 2015, proceedings, part III 18 (pp. 234-241). Springer international publishing.
[8] Schaefer, A. (2018). Local-global parcellation of the human cerebral cortex from intrinsic functional connectivity MRI. Cerebral cortex, 28(9), 3095-3114.
[9] Stanley, M. (2013). Defining nodes in complex brain networks. Frontiers in computational neuroscience, 7, 169.
[10] Tzourio-Mazoyer, N. (2002). Automated anatomical labeling of activations in SPM using a macroscopic anatomical parcellation of the MNI MRI single-subject brain. Neuroimage, 15(1), 273-289.