基于睡眠传感器的呼吸暂停严重程度预测数据

基于睡眠传感器采集的用户睡眠呼吸数据，建立呼吸暂停预测模型，计算呼吸暂停发生频次，用于评价用户的呼吸暂停严重程度。此数据主要用于用户的健康筛查服务，将呼吸暂停严重程度按发生频次的大小分为正常、轻度风险、中度风险、重度风险4个等级，当呼吸暂停严重程度较高时，及时提醒用户去医院检查，并且提供相关的健康建议和关怀服务。1、 预处理睡眠期间的呼吸特征序列数据，并将序列数据切分成多个固定时长的数据片段；
2、 将所有数据片段分别进行数学变换，如小波变换、傅里叶
变换、希尔伯特变换等；
3、 基于变换后的数据，进行特征提取，并通过特征筛查算法，筛选出重要特征，将特征送入ResNet深度神经网络模型进行训练和预测；
4、 模型输出结果为1，代表片段中存在呼吸暂停现象，模型输出结果为0，代表片段中不存在呼吸暂停现象；
5、 统计平均每小时呼吸暂停的发生次数，得到呼吸暂停指数；
6、 将呼吸暂停指数按从小到大划分成4个区间，分别代表正常、轻度风险、中度风险、重度风险，该4个区间由医学仪器多导仪同步实验后标定而来；
7、 根据呼吸暂停指数所在的数值区间，例如[0,5)、[5,15)、[15,25)、25及以上，分别代表正常、轻度风险、中度风险、重度风险。

This dataset is constructed using user sleep respiratory data collected by sleep sensors, with the goal of building apnea prediction models, calculating the frequency of apnea occurrences, and assessing the severity of sleep apnea in users. The data is mainly applied to user health screening services, where the severity of sleep apnea is divided into four grades—normal, mild risk, moderate risk, and severe risk—based on the frequency of apnea events. When the apnea severity is relatively high, users will be promptly reminded to visit a hospital for examination, and relevant health advice and care services will be provided.
1. Preprocess the respiratory feature sequence data collected during sleep, and split the sequence data into multiple fixed-duration data segments;
2. Perform mathematical transformations on all data segments, such as wavelet transform, Fourier transform, Hilbert transform, etc.;
3. Extract features from the transformed data, screen out important features via feature screening algorithms, and input the selected features into the ResNet deep neural network model for training and prediction;
4. A model output of 1 indicates that apnea is present in the segment, while an output of 0 indicates no apnea in the segment;
5. Count the average number of apnea occurrences per hour to obtain the apnea index;
6. Divide the apnea index into four intervals in ascending order, corresponding to normal, mild risk, moderate risk, and severe risk respectively. These four intervals are calibrated through synchronized experiments with medical polygraphs;
7. Determine the corresponding severity grade based on the interval where the apnea index falls. For example, [0, 5), [5, 15), [15, 25), and ≥25 correspond to normal, mild risk, moderate risk, and severe risk respectively.