A Short-Time Insomnia Detection System Based on Sleep EOG With RCMSE Analysis

Abstract

Objective: Currently, the prevalence of insomnia in 30% of the population in the word. To diagnose sleep issues, all-night polysomnography is usually taken from the patients, and the recordings are scored by a clinical staff. Nevertheless, manual sleep scoring and diagnosis are time consuming and subjective. In this study, a short-time insomnia detection system based on single-channel sleep EOG with refined composite multiscale entropy (RCMSE) analysis was proposed, and the performance of the proposed system was assessed with the manual scoring based on polysomnography. Methods: The sleep data from 32 subjects were used to develop and evaluate the proposed system; one half was healthy individual and the other half was insomnia patient. The corresponding RCMSE was computed from the short time single-channel sleep EOG (<; 30 min). Then, the mean values of their RCMSEs were computed. Finally, the mean values were used as input to an SVM classifier for insomnia detection. Results: 16 subjects were used to train a classifier; one half was healthy individual and the other half was insomnia patient; and the others were used to test. The averaged accuracy, sensitivity, specificity, kappa coefficient, and F1 score of the proposed system were 89.31%, 96.63%, 82.00%, 0.79, and 90.04%, respectively. Conclusion: Our results showed that RCMSE is a useful and representative feature for short-time insomnia detection. In addition, the proposed method has high accuracy and is good homecare applicability because a single-channel sleep EOG is used. Significance: In the future, we can integrate the proposed system with an EOG eye mask and portable PSG system for sleep quality assessment or insomnia screening in the home environment.

Keywords

• Short-time insomnia detection
• refine composite multiscale entropy
• single-channel sleep EOG
• support vector machine