A method based on independent component and support vector machine for the evaluation of mental status of sailors
Yinghua Liu, Weiming Zeng, Yingchao Shi, Nichuan Wang · Chin J Naut Med & Hyperbar Med · 2015
Objective To investigate a method based on independent component (IC) and support vector machine (SVM) for the detection of abnormal mental status of sailors. Methods First, the features of fingerprints were identified in 100 subjects from the control group and 88 subjects from the sailor group, and the twin support vector machine was used to establish the model for mental health evaluation. Then, the derrived evaluation model was used to evaluate the mental status of 88 sailors and identify those sailors with menatal disorder. Results Our research revealed that 5 out of 88 sailors were detected to be mentally abnormal. Further comparison of the default mode network between 5 abnormal sailors and that of the normal subjects revealed that there were significant differences between them. Conclusions With the prominent features of kurtosis, skewness, spatial entropy, degree of clustering, one-lag serial autocorrelation, temporal entropy and power contribution of the default mode network, the twin support vector machine could be used to establish the mental health evaluation model for the detection of those sailors with mental disorder. The research results could provide good objective standards for future evaluation of the mental health in sailors. Key words: Default mode network; Sailors; IC fingerprints; Support vector machine; Resting state