Mixtures of Dirichlet processes for joint spatial modelling of transcranial magnetic stimulation mapping data

• Published in: Journal of the Royal Statistical Society Series C: Applied Statistics Vol. 74; no. 1; pp. 1 - 17
• Main Authors: Egbon, Osafu Augustine, Heumann, Christian, Nascimento, Diego Carvalho, Louzada, Francisco
• Format: Journal Article
• Language: English
• Published: Oxford Oxford University Press 01.01.2025
• Subjects: Cortex; Mapping; Markov analysis; Markov chains; Mixtures; Monte Carlo simulation; Spatial analysis; Transcranial magnetic stimulation
• ISSN: 0035-9254; 1467-9876
• DOI: 10.1093/jrsssc/qlae042

A patient’s responses to Transcranial Magnetic Stimulation (TMS) pulses on the motor cortex have a complex spatial pattern, making it challenging to understand the response patterns across multiple patients. We developed a mixture of Dirichlet process models to distinguish between patient-specific and shared spatial patterns across multiple patients to provide insight into consistent response patterns essential for developing personalized treatment procedures. The Metropolis–Hastings within Gibbs sampler of the Markov Chain Monte Carlo algorithm was developed for estimation. The model was used to analyse the TMS data of 3 healthy subjects. The study revealed that the primary motor cortex of the hand consistently emerges as a promising region for eliciting optimal responses. This area serves as a key target for brain mapping using TMS to identify cortical hotspots. However, the excitability patterns in this region can vary significantly among patients.