基于口鼻流速信号与LSTM的实时睡眠呼吸暂停与低通气事件检测算法

An Apnea and Hypopnea Episodes Detection Algorithm Based on Oronasal Airflow Signals and LSTM Networks

睡眠呼吸暂停低通气综合征(Sleep Apnea-Hypopnea Syndrome,SAHS)不仅影响患者生活质量,而且可能诱发严重疾病,因此提出了一种基于口鼻流速信号与长短时记忆网络(Long Short-Term Memory,LSTM)的算法。该算法对于睡眠呼吸暂停及低通气事件的实时检测准确率为95%,与之前的研究相比,具有更高的准确率。由于本方法只需单通道信号,且具有实时性强的特点,可应用于家用无创呼吸机,简化SAHS的诊断和治疗流程。

Sleep Apnea-Hypopnea Syndrome(SAHS) affects patients’ quality of life and can induce serious illness.This study proposes an algorithm based on oronasal airflow signals and Long-Short-Term Memory(LSTM) neural networks to achieve real-time accurate detection of sleep apnea and hypopnea events with an accuracy rate of 95%.Compared with previous studies,this method has the characteristics of high accuracy,requiring only a single-channel signal,and strong real-time performance.Applying this method to home non-invasive ventilators can simplify the diagnosis and treatment process of SAHS.