Application of Hurst exponent in resting-state functional magnetic resonance imaging data

Objective Based on functional magnetic resonance imaging(fMRI) technology, the fractional Gaussian noise(fGn) model is adopted to study on the resting-state functional data. The Hurst exponent of fGn is used to characterize the BOLD signal. The distribution and test-retest reliability of Hurst exponent are analyzed. Methods Totally, 25 healthy adult were treated by resting-state f MRI scan for 3 times. MATLAB, DPARSF and REST software were used for data processing.The improved periodogram method was adopted to estimate Hurst exponent. The intra- class correlation coefficient(ICC)was introduced to perform statistical analysis. Results Significant differences were found in the Hurst exponent of different brain tissue. The Hurst exponent of gray matter was larger than 0.5, and the Hurst exponent of white matter was smaller than 0.5, while the Hurst exponent of cerebrospinal fluid was approximately 0.5. Based on the voxel or the average value of brain regions, the Hurst exponent showed satisfactory test- retest reliability. Conclusion fGn is suitable for the study on f MRI data. Hurst exponent can characterize the BOLD signal characteristics in many aspects, and Hurst exponent is helpful to further reveal the mechanism of brain activities.