The First Results of Systems Identification Methods for fMRI Data

The main goal of this paper is using identification methods as a certain alternative to DCM analysis which detects the so-called intrinsic connections among selected brain areas. In recent years it has been shown that the similar problems, as there appear in fMRI area, are formulated in dynamic system identification and estimation tasks. The subspace identification methods were chosen for this identification procedure because they prove good results for MIMO (Multiple Input Multiple Output) systems identification. These methods produce state space description of identified system. The main part of this paper deals with the quality of identification results depending on some important data parameters. Consequently the processing of final state space description into more suitable form follows.