Order detection for fMRI analysis: Joint estimation of downsampling depth and order by information theoretic criteria

Estimation of the order of functional magnetic resonance imaging (fMRI) data is a crucial step in data-driven methods assuming a multivariate linear model. Use of information theoretic criteria for model order detection was proven useful but the sample dependence in fMRI data limits this usefulness. In this paper, we propose an iterative procedure that jointly estimates the downsampling depth and order of fMRI data, both by using information theoretic criteria. Experimental results on real-world fMRI data show reliable performance of the new method. Order analysis on auditory oddball task (AOD) data of healthy and schizophrenia subjects suggests that model order can be a promising biomarker for mental disorders.