Abstract #2349

Decoding Task-Based Functional Connectivity: Exploring Kolmogorov-Arnold Networks for Task-Evoked fMRI Classification

Yuhan Chen¹, Zihao Tang^1,2,3, Tao Chen^2,4, Xinyi Wang^1,2, Yifei Sun⁵, Caleb The Tjoean¹, Jinglei Lv^2,5, Michael Barnett^2,6, Fernando Calamante^2,5,7, Muireann Irish^2,4, Weidong Cai¹, and Chenyu Wang^2,3,6

¹School of Computer Science, The University of Sydney, Sydney, Australia, ²Brain and Mind Centre, The University of Sydney, Sydney, Australia, ³Central Clinical School, The University of Sydney, Sydney, Australia, ⁴School of Psychology, The University of Sydney, Sydney, Australia, ⁵School of Biomedical Engineering, The University of Sydney, Sydney, Australia, ⁶Sydney Neuroimaging Analysis Centre, The University of Sydney, Sydney, Australia, ⁷Sydney Imaging, The University of Sydney, Sydney, Australia

Synopsis

Keywords: Functional Connectivity, fMRI (task based)

Motivation: Limited research exists on leveraging functional connectivity for task condition classification, and current deep learning models lack interpretability from a neuroscience perspective.

Goal(s): Evaluate the interpretability of Kolmogorov-Arnold Networks for task condition classification.

Approach: Kolmogorov-Arnold Networks was trained to classify task conditions using functional connectivity across seven tasks. Contribution analysis was conducted on learnable B-spline activation functions from correctly classified samples to report the important connections and regions.

Results: The analysis results from the trained Kolmogorov-Arnold Networks are more interpretable with respect to the literature on the functional roles of specific brain regions.

Impact: This study demonstrates that the Kolmogorov-Arnold Network offers strong interpretability for fMRI task-condition classification, aligning with literature on the functional roles of specific brain regions. This advances task-based fMRI analysis and supports the development of explainable neural networks in neuroscience.

How to access this content:

For one year after publication, abstracts and videos are only open to registrants of this annual meeting. Registrants should use their existing login information. Non-registrant access can be purchased via the ISMRM E-Library.

After one year, current ISMRM & ISMRT members get free access to both the abstracts and videos. Non-members and non-registrants must purchase access via the ISMRM E-Library.

After two years, the meeting proceedings (abstracts) are opened to the public and require no login information. Videos remain behind password for access by members, registrants and E-Library customers.

Click here for more information on becoming a member.