Spatiotemporal Gaussian Priors in Brain Source Reconstruction using Anisotropic Head Model: A study based on DTI-derived structural connectivity
Seneviratne, Gihani (2026)
2026
Master's Programme in Computing Sciences and Electrical Engineering
Informaatioteknologian ja viestinnän tiedekunta - Faculty of Information Technology and Communication Sciences
This publication is copyrighted. You may download, display and print it for Your own personal use. Commercial use is prohibited.
Hyväksymispäivämäärä
2026-06-05
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:tuni-202606036924
https://urn.fi/URN:NBN:fi:tuni-202606036924
Tiivistelmä
Electroencephalography (EEG) and magnetoencephalography (MEG) are non-invasive methods for studying dynamic neural activity; however, accurate source localization remains a challenging inverse problem due to the ill-posed nature of the reconstruction process, measurement noise, and the complexity of brain dynamics. Traditional inverse methods such as Minimum Norm Estimation (MNE) and standardized Low Resolution Electromagnetic Tomography (sLORETA) generally assume static source behavior and do not explicitly incorporate temporal evolution or anatomical brain connectivity.
This thesis investigates the incorporation of connectivity-informed spatiotemporal priors into the Standardized Kalman Filter (SKF) framework for dynamic EEG/MEG source reconstruction using an anisotropic finite element method (FEM) head model. Structural connectivity information derived from diffusion tensor imaging (DTI) streamline data was incorporated into the process noise covariance matrix through a weighted graph Laplacian and Gaussian Markov Random Field formulation and a regularization parameter (epsilon ϵ) was introduced to evaluate the process noise covariance matrix at different levels.
The proposed framework was implemented in the Zeffiro Interface environment using a multicompartment FEM head model based on the ICBM152 anatomical template. Synthetic Somatosensory Evoked Potential (SEP) signals were generated using deep thalamic and superficial postcentral cortical sources to evaluate the reconstruction performance under different epsilon values, Evolution Prior Signal-to-Noise Ratio (EP-SNR) values, smoothing conditions, and measurement noise levels.
Overall, the results showed that incorporating anatomical connectivity information into the SKF framework is effective in dynamic EEG/MEG source reconstruction. In particular, DTI-derived structural connectivity helped distinguish deep thalamic activity from superficial cortical activity under moderate noise conditions while maintaining stable reconstructions.
This thesis investigates the incorporation of connectivity-informed spatiotemporal priors into the Standardized Kalman Filter (SKF) framework for dynamic EEG/MEG source reconstruction using an anisotropic finite element method (FEM) head model. Structural connectivity information derived from diffusion tensor imaging (DTI) streamline data was incorporated into the process noise covariance matrix through a weighted graph Laplacian and Gaussian Markov Random Field formulation and a regularization parameter (epsilon ϵ) was introduced to evaluate the process noise covariance matrix at different levels.
The proposed framework was implemented in the Zeffiro Interface environment using a multicompartment FEM head model based on the ICBM152 anatomical template. Synthetic Somatosensory Evoked Potential (SEP) signals were generated using deep thalamic and superficial postcentral cortical sources to evaluate the reconstruction performance under different epsilon values, Evolution Prior Signal-to-Noise Ratio (EP-SNR) values, smoothing conditions, and measurement noise levels.
Overall, the results showed that incorporating anatomical connectivity information into the SKF framework is effective in dynamic EEG/MEG source reconstruction. In particular, DTI-derived structural connectivity helped distinguish deep thalamic activity from superficial cortical activity under moderate noise conditions while maintaining stable reconstructions.