Identification of essential language areas by combination of fMRI from different tasks using probabilistic independent component analysis

Functional magnetic resonance imaging (fMRI) has been used to lateralize and localize lan-guage areas for pre-operative planning pur-poses. To identify the essential language areas from this kind of observation method, we pro-pose an analysis strategy to combine fMRI data from two different tasks using probabilistic in-dependent component analysis (PICA). The assumption is that the independent compo-nents separated by PICA identify the networks activated by both tasks. The results from a study of twelve normal subjects showed that a language-specific component was consistently identified, with the participating networks sepa-rated into different components. Compared with a model-based method, PICA’s ability to capture the neural networks whose temporal activity may deviate from the task timing suggests that PICA may be more appropriate for analyzing language fMRI data with complex event-related paradigms, and may be particularly helpful for patient studies. This proposed strategy has the potential to improve the correlation between fMRI and invasive techniques which can dem-onstrate essential areas and which remain the clinical gold standard.