APNIWAVE: A Dataset Collected Using a Radar-Based Sleep-Apnea Screening Device for Use at Home

The "APNIWAVE: A Dataset Collected Using a Radar-Based Sleep-Apnea Screening Device for Use at Home" dataset was created by  collecting data from 11 patients with Obstructive Sleep Apnea and Hypopnea Syndrome (OSAHS) using an UWB sensor, which was the X4M200 UWB radar sensor by Novelda and a Raspberry Pi 3 device, for the purposes of APNIWAVE project (EIT Health RIS Scheme, Project ID: Project ID 2021-RIS_Innovation-066). The dataset comprises 1,011 selected samples of 10 sec each (a total of about 3 hours), from three main events that are the most frequent during sleep:

Normal breathing – periodic movement of the patient’s torso as a result of the inhale and exhale. (label in "Labels Three Events.csv" file equal to 0)

Apnea event – based on the Apnea type, e.g. central Apnea or hypopnea, no torso movement or a periodic torso movement but weaker than normal breathing, respectively. (label in "Labels Three Events.csv" file equal to 1)

Other event – events that may occur during data collection, such as change of posture, leaving the bedroom etc. (label in "Labels Three Events.csv" file equal to 2)

Data from five distances are included for each 10 sec sample. More specifically, the central distance is the distance of the user from the radar; on top of that we added two more discrete distance steps (as designated by the radar range resolution) in front of and another two behind the user, thus totaling five distances from the radar. These five distances were selected from a set of 165 distance steps (from 0.5m to 9 m) considering a distance step of about 0.05144 m and a sampling rate of 17 Hz. As a consequence, the size of each sample (of 10 sec duration) is 5 x 170, which leads to a total size of the uploaded "Multiple Distances Three Events.csv" file of 5 x 171870.

本数据集命名为APNIWAVE：一款基于雷达式睡眠呼吸暂停居家筛查设备采集的数据集，由APNIWAVE项目（EIT Health RIS Scheme，项目编号：2021-RIS_Innovation-066）研发团队构建。该数据集通过使用Novelda公司X4M200型超宽带（Ultra-Wideband, UWB）雷达传感器与树莓派3（Raspberry Pi 3）设备，对11名阻塞性睡眠呼吸暂停低通气综合征（Obstructive Sleep Apnea and Hypopnea Syndrome, OSAHS）患者开展数据采集工作。

数据集共筛选得到1011段时长为10秒的有效样本，总时长约3小时，涵盖睡眠过程中三类最常见事件：
1. 正常呼吸事件：患者因呼吸动作引发的躯干周期性运动，对应"Labels Three Events.csv"文件中的标签值为0；
2. 呼吸暂停/低通气事件：根据呼吸暂停类型（如中枢性呼吸暂停或低通气），分别表现为无躯干运动，或躯干周期性运动幅度弱于正常呼吸，对应"Labels Three Events.csv"文件中的标签值为1；
3. 其他事件：数据采集过程中可能出现的各类异常场景，例如体位改变、离开卧室等，对应"Labels Three Events.csv"文件中的标签值为2。

每段10秒样本均包含5个距离档位的采集数据。具体而言，中心档位为用户与雷达间的实际距离；在此基础上，依据雷达距离分辨率额外增设2个前方离散距离档位与2个后方离散距离档位，总计5个雷达距离档位。该5个档位从覆盖0.5m至9m范围的165个预设距离步长中选取，步长约0.05144m，采样率为17Hz。由此，每段10秒时长的样本数据维度为5×170，最终上传的"Multiple Distances Three Events.csv"文件总数据维度为5×171870。