Bayesian Spatial Process Models for Activation Patterns in Transcranial Magnetic Stimulation Mapping

• Autor(a) principal: Egbon, Osafu Augustine
• Data de Publicação: 2023
• Tipo de documento: Tese
• Idioma: eng
• Título da fonte: Biblioteca Digital de Teses e Dissertações da USP
• Texto Completo: https://www.teses.usp.br/teses/disponiveis/104/104131/tde-12092023-191817/
• topic: Brain mapping; Córtex Motor; Dirichlet process; Elicitação a Priori; Gaussian process; Mapeamento Cerebral; Motor cortex; Prior elicitation; Processo de Dirichlet; Processo Gaussiano

In recent years, Spatial statistical models have been gaining rapid attention for solving problems in biological systems due to the improvement in spatial data collection. It has proven extremely important in unveiling spatial patterns and predicting biological processes. This project developed novel parametric and nonparametric Bayesian spatial statistical models to analyze data generated by the muscular responses elicited by Transcranial magnetic stimulation (TMS) pulses induced on the motor cortex of a patient. The goal is to unveil new insights into patients response patterns important for achieving successful TMS therapy sessions. The first contribution of this project is a systematic review and meta-analysis of the existing Bayesian spatial models that could be considered for analyzing TMS datasets. The second contribution is the development of a user-friendly interface for performing Bayesian spatial modeling for analyzing TMS datasets based on state-of-the-art methods. The interface was documented in an R package, which is publicly available. The third contribution proposed novel spatial statistical models for integrating geostatistical datasets in the form of prior elicitation in a Bayesian analysis. The models were validated using simulation studies, and findings show that naively integrating geostatistical TMS datasets without ensuring the consistency of the data is detrimental to the desired inferences. The final contribution proposed a Bayesian nonparametric spatial model that leads to a non-stationary and non-Gaussian spatial process for the joint modeling of geostatistical TMS datasets. The method used a mixture of Dependent Dirichlet processes to share information across sub-spatial models. Two simulation studies were used to validate the model performance, and the result showed superior performance compared with independent and exchangeable models. The main finding of this work is that the primary motor cortex within the motor cortex region of the brain is responsible for the largest activation in the movement of the right first dorsal interosseous muscle. The finding also showed that the corticospinal excitability decreases with multiple TMS pulses on the motor cortex; however, it begins to regain its excitability strength after several stimulations. The findings from this project could guide TMS practitioners to improve patients treatment experiences.