论文部分内容阅读
睡眠分期是医学、神经信息领域的研究热点。人工标记睡眠数据是一项费时且费力的工作。自动睡眠分期方法能够减少人工分期的工作负荷,但在复杂多变的临床数据的应用上仍存在局限性。本文提出了一种改进的K均值聚类算法,主要目的是从实际睡眠数据的特点出发,研究睡眠自动分期方法。针对原始K均值聚类算法对初始聚类中心和离群点敏感的问题,本文结合密度的思想,选择周围数据密集的点作为初始中心,并根据“3σ法则”更新中心。改进算法在健康被试和接受持续正压通气(CPAP)治疗的睡眠障碍者的睡眠数据上进行了测试,平均分类精确度达到76%,同时结合实际睡眠数据的形态多样性验证讨论了该方法在临床数据上的可行性和有效性。
Sleep staging is a hot spot in the field of medicine and neural information. Manually marking sleep data is a time-consuming and laborious task. Automatic sleep staging can reduce the workload of manual staging, but there are still limitations in the application of complex clinical data. This paper presents an improved K-means clustering algorithm, the main purpose is to start from the characteristics of the actual sleep data to study sleep automatic staging method. Aiming at the problem that the original K-means clustering algorithm is sensitive to the initial clustering centers and outliers, this paper chooses the dense points around the data as the initial centers according to the idea of density and updates the centers according to the “3σ rule”. The improved algorithm was tested on sleep data of healthy subjects and sleep disorders treated with continuous positive airway pressure (CPAP), with an average classification accuracy of 76%. The method was also discussed based on morphological diversity of actual sleep data The clinical data on the feasibility and effectiveness.