多参数信息融合实现非脑电的睡眠结构分期

目前临床睡眠分析的主要手段是多导睡眠图（Polysomnograph,PSG）,用仪器记录被测者整晚的脑电、眼动电和颏肌电等生理参数,计算机进行自动睡眠分期,再由技术人员依据国际标准进行校正.几十年来PSG保持着睡眠分期金标准的地位.提出在不使用脑电的条件下,利用较易获得的心动周期、呼吸、体动等基本生理参数,提取其中与睡眠过程及其变化有关的规律和信息,建立知识规则库,采用不确定推理的证据理论进行多参数睡眠信息融合计算,实现睡眠结构分期.50余例与PSG对照试验结果表明：醒睡的平均符合率达90%以上,基本睡眠结构的平均符合率达75%以上,证明该技术达到了应用的要求.

A Non-EEG Approach to Sleep Analysis by Fusing Multi-Parameter Information

Objective: The traditional method of sleep architecture analysis based on polysomnography(PSG) contains electroencephalogram(EEG),electro-oculogram(EOG) and electromyogram(EMG),etc.The new approach proposed in this paper is free of EEG and other electrical physiological parameters,whose great advantage is exempting any need of direct attachment to the subject.Method: A patented mattress is used for all-night data acquisition.Parameters as heart beat interval,respiration period and body movement are related to sleep architecture.Through multi-resolution analysis of series of all-night heart-beat period and respiration period,a parameterized rule base is established to direct a inference process.Sleep architecture analysis is realized through fusing information in multi-parameter using Dempster evidence combination theory.Result: The method has been tested on 50 control experiment with PSG result as contrast,the wake-sleep coincidence rate is over 90%,wake-NREMS-REMS coincidence rate is larger tan 75%.The method is qualified as being useful in clinical application.