Enhanced Brain Source localization using Multimodal signal Fusion
Chaari, Anas
Chaari, Anas
Date
2025-12
Author
Advisor
Type
Thesis
Degree
Citations
Altmetric:
Files
35.232-2025.74a Anas Chaari_COMPRESSED.pdf
Adobe PDF, 2.35 MB
- Embargoed until 2027-05-21
Description
A Master of Science thesis in Biomedical Engineering by Anas Chaari entitled, “Enhanced Brain Source localization using Multimodal signal Fusion”, submitted in December 2025. Thesis advisor is Dr. Hasan Al Nashash and thesis co-advisor is Dr. Hasan Mir. Soft copy is available (Thesis, Completion Certificate, Approval Signatures, and AUS Archives Consent Form).
Abstract
The human brain is the most complex organ that controls multiple cognitive, sensory, and motor functions. Understanding its complex dynamics requires precise brain source localization techniques, which are important for diagnosing neurological disorders and studying brain functions. Multimodal neuroimaging can be applied to improve localization accuracy by combining modalities with complementary strengths of each modality. This project proposes an approach for multimodal brain source localization by integrating electroencephalography (EEG) and functional near-infrared spectroscopy (fNIRS) data using Restricted Maximum Likelihood (REML) model to enhance both spatial and temporal accuracy. The methodology is validated using simulated data, where known neural sources were reconstructed independently using EEG and fNIRS, and then jointly using fNIRS-derived spatial priors. Integration reduced mean localization error (MLE) from 103.46 mm to 22.75 mm, showing the method’s accuracy. The proposed methodology was then applied to experimental data from 20 healthy subjects performing cognitive stress-induction tasks. EEG processing included filtering and ICA-based artifact removal, while fNIRS data underwent motion correction and detrending. The results showed consistent improvements across detection performance metrics, including specificity, accuracy, and MLE. For example, in Subject 10, MLE reduced from 72.601 mm (EEG only) to 16.010 mm (integrated), alongside improvements in specificity (0.3359 to 0.9883) and accuracy (0.3373 to 0.9873). A novel contribution of this work is the inclusion of neurovascular coupling delay compensation prior to multimodal integration, where EEG data are time shifted relative to hemodynamic responses in 2 s increments from 3 to 21 s. Optimal alignment was achieved between 3–5 s, showing enhanced localization performance. Additionally, analysis using frontal-only EEG channels to match the fNIRS cap layout im- proved spatial constraint and reduced interference from unrelated cortical regions.