AN EM ALGORITHM FOR RICIAN FMRI ACTIVATION DETECTION

In functional magnetic resonance imaging (IMRI), most of studies to detect activations in brain are using magnitude time courses which in fact obey Rice distribution. When signal to noise ratio (SNR) of a time course is high, it is known that the Gaussian approximation of Rice distribution works well. However, SNRs are not known before data analysis and imaging with high spatial resolution which decreases SNRs is increasingly required in functional studies. In this paper, we suggest a method to build up an activation map based on Rician distributed modeling via expectation maximization (EM) algorithm. This Rician-EM allows simple iterations and very easy interpretations related to an existing approach. We perform simulations to compare the developed technique with an existing approach.