论文部分内容阅读
睡眠的分级研究是睡眠状况分析和睡眠质量评价的前提和基本内容。目前国际通用的睡眠分级方法 ,是利用脑电信号另加脑功能信号 (如肌电图、眼动电流图 ) ,且必须由人工判别分析。大脑皮层互信息理论是研究脑功能变化的有力工具。通过动态计算睡眠脑电4个导联之间的互信息时间序列的复杂度 ,并利用一个 3层的人工神经网络进行 6个级别的分类。 6例 72 0个不同时期的睡眠片段的测试表明 ,系统睡眠分级与人工分级的总相符率达到 90 .83 % ,且实现了睡眠动态自动分级。神经网络的学习功能 ,可使系统的准确率进一步提高 ,逐渐接近或达到人工分级的水平。与其他睡眠分级方法相比 ,本系统有一定优势 ,且计算速度快 ,可望应用于临床实时睡眠监护及睡眠分析中。
The classification of sleep is the prerequisite and basic content of sleep condition analysis and sleep quality evaluation. At present, the internationally accepted sleep classification method is to use EEG signals plus brain function signals (such as EMG and EKG) and must be manually discriminated and analyzed. The theory of cerebral cortex mutual information is a powerful tool to study the changes of brain function. By dynamically calculating the complexity of the time series of mutual information between four leads of sleep-induced EEG and using a 3-layer artificial neural network to perform 6-level classification. The test of 6 sleep segments in 7200 different periods showed that the total coincidence rate of system sleep classification and artificial classification reached 90.83%, and the sleep dynamic automatic classification was realized. The learning function of neural network can further improve the accuracy of the system and gradually approach or reach the level of manual grading. Compared with other sleep classification methods, the system has certain advantages, and the calculation speed, is expected to be used in clinical real-time sleep monitoring and sleep analysis.