Standardized Kalman filtering for dynamical source localization of concurrent subcortical and cortical brain activity

• Published in: Clinical neurophysiology 2024-12, Vol.168, p.15-24
• Main Authors: Lahtinen, Joonas, Ronni, Paavo, Subramaniyam, Narayan Puthanmadam, Koulouri, Alexandra, Wolters, Carsten, Pursiainen, Sampsa
• Format: Article
• Language: English
• Subjects: Bayes Theorem; Brain - physiology; Brain Mapping - methods; Cerebral Cortex - physiology; Electroencephalography (EEG); Electroencephalography - methods; Electroencephalography - standards; Evoked Potentials, Somatosensory - physiology; Humans; Inverse methods; Kalman filtering; Source localization; Spatiotemporal; Standardization
• ISSN: 1388-2457; 1872-8952; 1872-8952
• DOI: 10.1016/j.clinph.2024.09.021

•Spatiotemporal standardization is proposed for the Kalman filter.•Standardized Kalman filter (SKF) tracked better than compared methods in simulations.•SKF localizes known originators of SEP accurately with numerical and real data. We introduce standardized Kalman filtering (SKF) as a new spatiotemporal method for tracking brain activity. Via the Kalman filtering scheme, the computational workload is low, and by spatiotemporal standardization, we reduce the depth bias of non-standardized Kalman filtering (KF). We describe the standardized KF methodology for spatiotemporal tracking from the Bayesian perspective. We construct a realistic simulation setup that resembles activity due to somatosensory evoked potential (SEP) to validate the proposed methodology before we run our tests using real SEP data. In the experiments, SKF was compared with standardized low-resolution brain electromagnetic tomography (sLORETA) and the non-standardized KF. SKF localized the cortical and subcortical SEP originators appropriately and tracked P20/N20 originators for investigated signal-to-noise ratios (25, 15, and 5 dB). sLORETA distinguished those for 25 and 15 dB suppressing the subcortical originators. KF tracked only the evolution of cortical activity but mislocalized it. The numerical results suggest that SKF inherits the estimation accuracy of sLORETA and traceability of KF while producing focal estimates for SEP originators. SKF could help study time-evolving brain activities and localize landmarks with a deep contributor or when there is no prior knowledge of evolution.