Uncertainties Study Of FMRI Data Processing Based On ReHo Algorithm

Functional magnetic resonance imaging(fMRI) is a non-invasive, repeatable and lower-cost functional brain imaging technology, used to identify the brain function of spontaneous activity. Regional homogeneity(Re Ho) algorithm is a driven analysis method which is widely applied in the research on schizophrenia, depression and other serious brain disease. There are some uncertainties in the process of Re Ho fMRI data analysis, which will affect the reliability of the results. Some scholars have carried out some research on the uncertainty of the preprocess parameters, data standardization method and so on. However, these research usually only pay attention to the influence of Re Ho value, but no analysis of the effect of abnormal brain regions. Therefore, it is needed to study the effect of these uncertainties on the results of fMRI data analysis based on Re Ho algorithm.In this paper, the dataset is based on some schizophrenia with and without agression. The Re Ho algorithm is applied for each patient and then statistical analysis shows the differences between the two groups of the brain. In this foundation, a quantitative method is proposed to describe the similarity of the results, and to evaluate the incidence of uncertainties in the Re Ho algorithm.The main work and conclusions in this paper are as follows:(1) The basic principle and optional parameters of fMRI data processing are analyzed and the uncertainties in the process of fMRI data preprocessing, Re Ho algorithm and statistical analysis are found out.(2) Dice similarity coefficient(DC) is proposed and the method of contribution rate evaluation of experimental results in each group. Including slice timing, a standard template selection, linear drifting, frequency domain filtering, standardization method, spatial smoothing kernel size and multiple comparison correction method, the seven the uncertainties which existed in the data processing and the statistical analysis have been conducted the contrast researches. By contrast analysis, Re Ho algorithm of the signal frequency domain of the Band Select the most sensitive, the data processing of the process will have to pay more atte tion to signal frequency band selection. In addition to the spatial smoothing kernel will also cause greater difference on the results, the statistical analysis of the results are easily affected. Moreover, linear drifting has the smallest effect on experimental results.(3) The intraclass correlation coefficient was used to obtain the relationship among the time series with different length data which calculated under the Re Ho algorithm and standards. Experimental results show that the length of time series is 3 minutes or more have a better reliability of the calculated value in Re Ho algorithm, it is similar to the results of the research in the literatures. In addition, the effect of fMRI data of different time length on the results of abnormal brain regions was analyzed, the results show that the longer length of time the more abnormal brain be found, and the abnormal of the occipital gyrus is found whether in the long or short time.(4) The influence of three different neighborhood sizes of Re Ho algorithm is analyzed. The more the neighborhood is, the more abnormal brain regions are, and the results of the smaller neighborhood are almost entirely of the results of the larger neighborhood.