Atena Rezaei: Utilizing novel mathematical techniques for detecting brain activity via recently developed interface
Detecting the brain activity, particularly, weakly distinguishable deep activity is significant for the neuroscience community for the treatment of brain disorders such as Parkinson’s disease, Alzheimer’s disease, or refractory epilepsy. To this end, we focused on developing and applying inverse modelling for robust source localization of the neural activity from non-invasive measurement. However, the solution of the inverse approach is heavily dependent on the accuracy of the associated forward technique. Our findings reveal that the finite element method (FEM)-based forward approach employed in this study is accurate enough to model the electromagnetic field of the brain compared to existing classical forward techniques, especially in the case of thin cortices such as children or pathological applications. Furthermore, the inverse modelling techniques in the framework of the hierarchical Bayesian modelling (HBM) have been developed to reconstruct the generators of EEG and MEG signals, i.e., primary current density of neural activity from different depths of the brain. Our inverse approaches can detect the far-field activity, which is known to be challenging since the subcortical domain is far from the sensors located on the scalp. In our investigation, the proposed approaches are implemented and integrated in Zeffiro Interface (ZI), which is an openly available MATLAB toolbox for forward and inverse computations. ZI uses a realistic and heavily folded head model while utilizing graphics processing unit (GPU) to accelerate the computations and minimize the computing cost with standard computers or laptops.
In our study, the somatosensory cortex (SI) as cortical area and thalamus and brainstem as corresponding subcortical areas are considered since they are involved in sensory processing. In our experiments, the datasets of somatosensory evoked potentials (SEPs) and somatosensory evoked fields (SEFs) of the median nerve stimulation are utilized to localize the P20/N20 component corresponding the somatosensory evoked responses at posterior wall of the central sulcus, Brodmann area 3b at 20ms. Additionally, we investigate analyzing thalamocortical connections at consecutive latencies, i.e., 14-30ms. Following this, our developed inverse approach, randomized multiresolution scanning (RAMUS), is employed to detect the P14/N14 component at medial lemniscus pathway at 14ms for the earliest latency. Our results also show that simultaneous cortical and subcortical activity can be detected at 20ms at Brodmann area 3b and VPL (ventral posterolateral) thalamus, respectively. In addition, in this dissertation we proposed a model for optimization and parameters’ selection of HBM-based inversion techniques.
“Our brain is a complex and mysterious organ. Applying and investigating novel mathematical techniques can open new windows toward detecting brain activity at different depths and help medical doctors to better understand the functionality of the brain for treatment of brain disorders such as epilepsy.” says MSc Atena Rezaei.
The doctoral dissertation of MSc (Tech) Atena Rezaei in the field of Applied Mathematics titled Forward and Inverse Modelling via Finite Elements in EEG/MEG Source Localization: Application to Event Related Responses will be publicly examined in the Faculty of Engineering and Natural Sciences at Tampere University at 12 o’clock on Friday 17.12.2021 at S2 auditorium of the Sähkötalo, Hervanta campus, Tampere University. The Opponent will be Associate Professor Stephanie Lohrengel from Universite de Reims Champagne-Ardenne. The Custos will be Associate Professor Sampsa Pursiainen from the Faculty of Information Technology and Communication Sciences, Tampere University.
The public can attend the event via Panopto remote connection.
The dissertation is available online at the http://urn.fi/URN:ISBN:978-952-03-2180-2