The Effect of Prior Parameters on Standardized Kalman Filter-Based EEG/MEG Source Localization: A simulation study based on somatosensory evoked potentials
Wanni Achchi Kankanamge, Dilshanie Prasikala (2024)
2024
Tietojenkäsittelyopin maisteriohjelma - Master's Programme in Computer Science
Informaatioteknologian ja viestinnän tiedekunta - Faculty of Information Technology and Communication Sciences
Hyväksymispäivämäärä
2024-12-10
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:tuni-2024120810843
https://urn.fi/URN:NBN:fi:tuni-2024120810843
Tiivistelmä
EEG/MEG source localization is an important branch in neuroscience as it has many practical applications like epilepsy diagnosis and treatment planning. Source localization consists of forward and inverse problems where the inverse problem is ill-posed and non-uniquely solvable. Prior knowledge is needed to arrive at a unique solution. Employing the Bayesian perspective makes it possible to incorporate prior knowledge into source localization algorithms.
Many source localization algorithms are currently utilized to solve this problem and they have their own strengths and weaknesses. There are a limited number of algorithms that take the spatiotemporal nature of data into consideration. Kalman filtering algorithm is one such algorithm that considers the dynamic nature of brain signals. But it suffers from depth bias.
Standardized Kalman filtering algorithm(SKF) uses a post-hoc weighting approach that has proven to overcome this depth bias. However, this algorithm is not been extensively studied in the literature. In the SKF algorithm, prior knowledge is induced as prior parameters.
The main objective of this study was to investigate the effect of prior parameters on the quality of reconstruction using synthetic data corresponding to SEP signals. Other than the main objective, it was expected to investigate the possibility of improving the reconstruction by introducing a novel approach to modify the prior parameters with additional knowledge. The results from the SKF algorithm were compared to IAS algorithm at three different noise levels.
From the results, rough values for prior parameters to obtain a good reconstruction for the data were discovered. In addition, the SKF algorithm was found to be more robust to background noise than the IAS algorithm. More studies need to be carried out to find the possibility of inducing additional information to the prior parameters.
Many source localization algorithms are currently utilized to solve this problem and they have their own strengths and weaknesses. There are a limited number of algorithms that take the spatiotemporal nature of data into consideration. Kalman filtering algorithm is one such algorithm that considers the dynamic nature of brain signals. But it suffers from depth bias.
Standardized Kalman filtering algorithm(SKF) uses a post-hoc weighting approach that has proven to overcome this depth bias. However, this algorithm is not been extensively studied in the literature. In the SKF algorithm, prior knowledge is induced as prior parameters.
The main objective of this study was to investigate the effect of prior parameters on the quality of reconstruction using synthetic data corresponding to SEP signals. Other than the main objective, it was expected to investigate the possibility of improving the reconstruction by introducing a novel approach to modify the prior parameters with additional knowledge. The results from the SKF algorithm were compared to IAS algorithm at three different noise levels.
From the results, rough values for prior parameters to obtain a good reconstruction for the data were discovered. In addition, the SKF algorithm was found to be more robust to background noise than the IAS algorithm. More studies need to be carried out to find the possibility of inducing additional information to the prior parameters.