SUMMARY
In the last decades, magnetic resonance imaging gained lot of popularity, and also functional magnetic resonance imaging (fMRI), due to the fact that MRI is a harmless and efficient technique for human cerebral activity studies; fMRI aims to determine and to locate different brain activities when the subject is doing a predetermined task. In addition, using fMRI analysis, nowadays we can make prediction on several diseases. This paper’s purpose is to describe the General Linear Model for fMRI statistical analysis algorithm, for a 64 x 64 x 22 voxels dataset on a revolutionary parallel computing machine, Connex Array. We make a comparison to other computing machines used in the same purpose, in terms of algorithm time execution (statistical analysis speed). We will show that by taking advantage on its specific parallel computation each step in GLM analysis, Connex Array is able to answer successfully to computational challenge launched by fMRI computation: thespeed-up.