Affiliations: [a] College of Computer Science and Engineering, Shandong University of Science and Technology, Qingdao, China
| [b]
The Affiliated Hospital of Qingdao University, Qingdao, China
| [c] Provincial Key Lab. for Information Technology of Wisdom Mining of Shandong Province, Shandong University of Science and Technology, Qingdao, China
Abstract: As a common disease, migraine has a high incidence but the pathogenesis is still not clear. Resting-state Functional Magnetic Resonance Imaging (rs-fMRI) is an important research topic in the field of brain medicine, which can classify rs-fMRI data to automatically diagnose brain diseases. However, the original features of the rs-fMRI data are difficult to be extracted and the high-dimensional characteristics, which make the data analysis an extremely complicated task. Those have also plagued many researchers and bring great challenges to the existing pattern classification methods. Aiming at the high dimensionality of rs-fMRI data, in this paper, we propose a feature extraction approach based on the combination of neighborhood rough set and PCA, thereby improving the accuracy of migraine identification. Firstly, Resting-State fMRI Data Analysis Toolkit plus was applied for preprocessing, calculating three characteristic indices: Amplitude of Low Frequency Fluctuation (ALFF), Regional Homogeneity (ReHo) and Functional Connectivity (FC. The inter-group difference analysis was performed by two-sample T test and GRF correction. Then, correlation coefficient matrix original features extraction was performed by means of automatic anatomical label template (AAL). Finally, the original features were trained by the traditional classification algorithm in machine learning. The experimental results show that the propose approach can obtain good performance in predicting migraine.
Keywords: Resting state functional magnetic resonance imaging, migraine, correlation coefficient matrix, neighborhood rough set, principal component analysis, classification
