Hassan Ahnaf Rashik, Bhuiyan Mohammed Imamul Hassan
Department of Electrical and Electronic Engineering, Bangladesh University of Engineering and Technology, Dhaka 1205, Bangladesh.
Department of Electrical and Electronic Engineering, Bangladesh University of Engineering and Technology, Dhaka 1205, Bangladesh.
Comput Methods Programs Biomed. 2017 Mar;140:201-210. doi: 10.1016/j.cmpb.2016.12.015. Epub 2016 Dec 27.
Automatic sleep staging is essential for alleviating the burden of the physicians of analyzing a large volume of data by visual inspection. It is also a precondition for making an automated sleep monitoring system feasible. Further, computerized sleep scoring will expedite large-scale data analysis in sleep research. Nevertheless, most of the existing works on sleep staging are either multichannel or multiple physiological signal based which are uncomfortable for the user and hinder the feasibility of an in-home sleep monitoring device. So, a successful and reliable computer-assisted sleep staging scheme is yet to emerge.
In this work, we propose a single channel EEG based algorithm for computerized sleep scoring. In the proposed algorithm, we decompose EEG signal segments using Ensemble Empirical Mode Decomposition (EEMD) and extract various statistical moment based features. The effectiveness of EEMD and statistical features are investigated. Statistical analysis is performed for feature selection. A newly proposed classification technique, namely - Random under sampling boosting (RUSBoost) is introduced for sleep stage classification. This is the first implementation of EEMD in conjunction with RUSBoost to the best of the authors' knowledge. The proposed feature extraction scheme's performance is investigated for various choices of classification models. The algorithmic performance of our scheme is evaluated against contemporary works in the literature.
The performance of the proposed method is comparable or better than that of the state-of-the-art ones. The proposed algorithm gives 88.07%, 83.49%, 92.66%, 94.23%, and 98.15% for 6-state to 2-state classification of sleep stages on Sleep-EDF database. Our experimental outcomes reveal that RUSBoost outperforms other classification models for the feature extraction framework presented in this work. Besides, the algorithm proposed in this work demonstrates high detection accuracy for the sleep states S1 and REM.
Statistical moment based features in the EEMD domain distinguish the sleep states successfully and efficaciously. The automated sleep scoring scheme propounded herein can eradicate the onus of the clinicians, contribute to the device implementation of a sleep monitoring system, and benefit sleep research.
自动睡眠分期对于减轻医生通过目视检查分析大量数据的负担至关重要。它也是使自动睡眠监测系统可行的前提条件。此外，计算机化睡眠评分将加快睡眠研究中的大规模数据分析。然而，现有的大多数睡眠分期工作要么基于多通道，要么基于多种生理信号，这对用户来说不太舒适，并且阻碍了家用睡眠监测设备的可行性。因此，一个成功且可靠的计算机辅助睡眠分期方案尚未出现。
在这项工作中，我们提出了一种基于单通道脑电图的计算机化睡眠评分算法。在所提出的算法中，我们使用总体经验模态分解（EEMD）对脑电图信号段进行分解，并提取基于各种统计矩的特征。研究了EEMD和统计特征的有效性。进行统计分析以进行特征选择。引入了一种新提出的分类技术，即随机欠采样增强（RUSBoost）用于睡眠阶段分类。据作者所知，这是首次将EEMD与RUSBoost结合使用。针对各种分类模型的选择，研究了所提出的特征提取方案的性能。我们方案的算法性能与文献中的当代作品进行了比较。
所提出方法的性能与现有技术相当或更好。在所提出的算法中，对于Sleep-EDF数据库上睡眠阶段从6状态到2状态的分类，分别给出了88.07%、83.49%、92.66%、94.23%和98.15%的准确率。我们的实验结果表明，对于本文提出的特征提取框架，RUSBoost优于其他分类模型。此外，本文提出的算法对睡眠状态S1和快速眼动（REM）表现出较高的检测准确率。
EEMD域中基于统计矩的特征能够成功且有效地区分睡眠状态。本文提出的自动睡眠评分方案可以消除临床医生的负担，有助于睡眠监测系统的设备实现，并有益于睡眠研究。
