计算机辅助的意识障碍患者诊断方法研究

In this paper, support vector machine (SVM) was performed on patients with disorder of consciousness (DOC) based on diffusion kurtosis imaging (DKI) and white matter fiber tracking. Firstly, the DKI parameters were calculated by atlas-based method, and the diffusion orientation distribution function (dODF) was estimated according to the kurtosis tensor and diffusion tensor, and then tracking the fiber bundles passing through the region of interest (ROI) based on dODF to obtain the parameters of fiber bundle. Secondly, the feature extraction was carried out by independent sample t test, and the age and gender were added separately as a feature to train the classifier. Results show that the classifier get the highest accuracy with the age as a feature, reaching 91.67%.And the performance of the classifier is further improved while introducing the virtual sample generation technique based on the Gaussian distribution. The accuracy of the cross validation is 100%, and the prediction accuracy of the new sample is 92.59%. The results show that the classification based on DKI and fiber bundle tracking can be used as a reference for the diagnosis of DOC patients.