基于共振峰的OSAHS筛查

发布时间：2018-10-08 14:54

【摘要】：阻塞性睡眠呼吸暂停低通气综合症是一种发病率很高的睡眠呼吸紊乱性疾病,睡眠期间的频繁呼吸暂停和低通气使得患者更容易引发心血管疾病、高血压、肾脏疾病等其他生命器官并发症,甚至发生猝死。多导睡眠监测被公认为是诊断睡眠呼吸障碍疾病的“金标准”,但是由于多导睡眠监测设备有限、检测费用昂贵、监测过程不舒适等缺点导致大部分打鼾者得不到及时的诊断。目前迫切需要找到一种便携的、舒适的、低费用的、可用于大量人群的筛查方法来减轻多导睡眠监测的负荷。本文研究方法是利用鼾声信号的共振峰参数来实现阻塞性睡眠呼吸暂停低通气综合症的筛查。首先利用数字语音信号处理的方法对鼾声信号进行预处理,利用一种改进的基于短时能量的方法检测出所有的鼾声段语音；利用线性预测技术估计产生鼾声的上气道模型参数,并利用求根方法计算出鼾声段的第一共振峰频率。目前已有方法使用统一固定的共振峰阈值来区分正常鼾声段和不正常鼾声段,但是每个人的上气道生理结构是不同的,即存在个体差异,现有的固定共振峰门限值筛查方法受个体差异的影响存在筛查率不高的缺陷；本文方法提出了一种不受个体差异影响的个体化阈值,利用K均值聚类算法将打鼾者一整晚鼾声段的第一共振峰频率分为两类,并将较小的聚类中心(正常鼾声对应的第一共振峰频率)视为该打鼾者的基准频率,根据基准频率与不正常鼾声第一共振峰频率的关系得到个体化阈值。本文提出阻塞性睡眠呼吸暂停低通气综合症筛查方法的依据为：首先,如果第一共振峰频率值高于个体化阈值,就认为是不正常鼾声段的第一共振峰频率；如果不正常鼾声段的持续时间大于0.3s,则认为鼾声段是不正常鼾声段；其次,模拟多导睡眠监测的呼吸暂停—低通气指数(AHI),即统计一小时内不正常鼾声段的个数,根据多导睡眠监测的标准,如果AHI高于5次/时,则就初步认为该打鼾者是阻塞性睡眠呼吸暂停低通气综合症患者,否则认为是单纯的打鼾者。本文筛查方法的灵敏度和特异度分别是93.3%和91.67%,满足临床医学上筛查疾病的要求。
[Abstract]:Obstructive sleep apnea hypopnea syndrome (OSAS) is a high incidence of sleep apnea disorder disease. Frequent apnea and hypopnea during sleep make patients more prone to cardiovascular disease and hypertension. Kidney disease and other life organ complications, and even sudden death. Polysomnography is recognized as the "golden standard" for the diagnosis of sleep apnea disorder. However, due to the limitation of polysomnography monitoring equipment, the high cost of detection and the discomfort of monitoring process, most snorers can not be diagnosed in time. There is an urgent need to find a portable, comfortable, low-cost screening method that can be used in large populations to reduce the load of polysomnography. In this paper, the resonant peak parameters of snoring signal are used to screen obstructive sleep apnea hypopnea syndrome (OSAS). Firstly, the snoring signal is preprocessed by digital speech signal processing, and all snoring segment speech is detected by an improved method based on short time energy, and the parameters of upper airway model which produce snoring are estimated by linear prediction technique. The first resonance peak frequency of snoring is calculated by root seeking method. At present, there are methods to distinguish normal snoring segment from abnormal snoring segment by using a fixed resonance peak threshold, but the physiological structure of upper airway is different, that is, individual differences exist. The existing screening methods with fixed resonance peak threshold have the defect that the screening rate is not high due to individual differences. In this paper, an individual threshold is proposed, which is not affected by individual differences. K-means clustering algorithm is used to divide the first resonance peak frequency of snoring all night into two categories, and the smaller cluster center (the first resonance peak frequency corresponding to normal snoring) is regarded as the reference frequency. According to the relation between the reference frequency and the frequency of the first resonance peak of abnormal snoring, the individualized threshold is obtained. In this paper, a screening method for obstructive sleep apnea hypopnea syndrome (OSAS) is proposed. Firstly, if the frequency of the first resonance peak is higher than the individual threshold, it is considered to be the first resonance peak frequency of the abnormal snoring segment. If the duration of abnormal snoring segment is longer than 0.3 s, the snoring segment is considered to be abnormal snoring segment. Secondly, the apnea hypopnea index (AHI), which simulates polysomnotic monitoring, counts the number of abnormal snoring segments within an hour. According to the standard of polysomnography, if AHI is more than 5 times / time, the snoring person is considered as obstructive sleep apnea hypopnea syndrome, otherwise it is considered as a simple snoring person. The sensitivity and specificity of this screening method are 93.3% and 91.67% respectively, which meet the requirements of clinical medical screening.
【学位授予单位】：大连理工大学
【学位级别】：硕士
【学位授予年份】：2011
【分类号】：R766

【参考文献】