Optimal Analysis of Functional Magnetic Resonance (fMRI) Using Nonparametric Permutation Method

Abstract

The non-parametric permutation approach is an intuitive and flexible approach for the statistical analysis of fMRI data compared to the parametric techniques. This approach can also be used to verify the validity of less computationally expensive parametric approaches. The methodology concept and comparative features of non-parametric permutation methods have been handled and discussed by various researchers. However, no available explication of the method exists, also, no freely distributed program is implemented. Consequently, this technique has not been applied practically. In the current work, the App Designer Statistical Parametric and Non-parametric Mapping of Functional Magnetic Resonance Imaging (SPnPM fMRI) tool is proposed to address these issues. The SPnPM fMRI toolbox is an open-source package that is designed for crucial comparison analysis. The comparison is performed between the statistical parametric and non-parametric mapping of the second-level analysis of fMRI data. T-test utilizing Random Field Theory (RFT), smoothed pseudo-t-test applying a permutation test, and t-statistic utilizing a permutation test without smoothing, are carried out. In addition, the corresponding parametric results are qualitatively and quantitatively compared. The outcomes on real fMRI data show the non-parametric approach achieved suprathreshold voxels more than those achieved in the conventional approach. Such as the activation of the anterior cingulate at (3, 15, 45) coordinates equal to 402 voxels employing a non-parametric pseudo-t-test, while 75, 28 voxels at the same coordinates using the T-test non-parametric and t-test parametric using random field theory, respectively. Besides, more suitable for small sample sizes.