Accurate autocorrelation modeling substantially improves fMRI reliability

Abstract

Given the recent controversies in some neuroimaging statistical methods, we compared the most frequently used functional Magnetic Resonance Imaging (fMRI) analysis packages: AFNI, FSL and SPM, with regard to temporal autocorrelation modeling. This process, sometimes known as pre-whitening, is conducted in virtually all task fMRI studies. We employed eleven datasets containing 980 scans corresponding to different fMRI protocols and subject populations. Though autocorrelation modeling in AFNI was not perfect, its performance was much higher than the performance of autocorrelation modeling in FSL and SPM. The residual autocorrelated noise in FSL and SPM led to heavily confounded first level results, particularly for low-frequency experimental designs. Also, we observed very severe problems for scans with short repetition times. The resulting false positives and false negatives can be expected to propagate to the group level, especially if the group analysis is performed with a mixed effects model. Our results show superior performance of SPM's alternative pre-whitening: FAST, over the default SPM's method. The reliability of task fMRI studies would increase with more accurate autocorrelation modeling. Furthermore, reliability could increase if the analysis packages provided diagnostic plots. This way the investigator would be aware of residual autocorrelated noise in the GLM residuals. We provide a MATLAB script for the fMRI researchers to check if their analyses might be affected by imperfect pre-whitening.