Sparse-KSVD for Blind Decomposition on Task-Related fMRI Data: A Comparative Analysis with ICA and MCA-KSVD
Nguyen P, Nguyen PK (2025)
Publication Type: Conference contribution
Publication year: 2025
Event location: Dresden, Sachsen
Journal Issue: 2025
DOI: 10.1109/MOCAST65744.2025.11083729
Abstract
Functional MRI (fMRI) data analysis enables researchers to partition the brain into distinct regions, offering valuable insights into neurological disorders. A widely used decomposition technique is Independent Component Analysis (ICA), which identifies functional brain networks by assuming statistical independence between components. However, this assumption may not fully align with the physiological characteristics of fMRI data. To address these limitations, this study explores Sparse-KSVD, a KSVD-based method with limited prior application in fMRI research, for identifying functional brain networks. Sparse-KSVD is systematically compared to MCA-KSVD, another KSVD-family method, and the conventional ICA using task-related fMRI data. The analysis shows that Sparse-KSVD accurately identifies functional brain networks with results comparable to those of ICA and MCA-KSVD. Furthermore, Sparse-KSVD outperforms in execution time, signal localization, and spatial resolution under optimal parameter settings. These findings demonstrate that Sparse-KSVD is a promising alternative to ICA and MCA-KSVD. However, careful parameter tuning is essential to achieve its full potential.
Authors with CRIS profile
Involved external institutions
How to cite
APA:
Nguyen, P., & Nguyen, P.K. (2025). Sparse-KSVD for Blind Decomposition on Task-Related fMRI Data: A Comparative Analysis with ICA and MCA-KSVD. In IEEE (Eds.), Proceedings of the 14. Internationale Konferenz über moderne Schaltungs- und Systemtechnologien (MOCAST). Dresden, Sachsen, DE.
MLA:
Nguyen, Phuc, and Phuoc Khang Nguyen. "Sparse-KSVD for Blind Decomposition on Task-Related fMRI Data: A Comparative Analysis with ICA and MCA-KSVD." Proceedings of the 14. Internationale Konferenz über moderne Schaltungs- und Systemtechnologien (MOCAST), Dresden, Sachsen Ed. IEEE, 2025.
BibTeX: Download