Blood Saturation Decreasing Level Based on the Features of a Spirogram Signal

Obstructive sleep apnea is a common disease associated with respiratory processes during sleep due to the narrowing of the upper respiratory airways, which leads to a decrease in the amount of oxygen in arterial blood flow - SpO2 - to human internal organs. Low blood saturation values can lead to serious consequences in the patient's health, and it can be monitored, for example, using pulsoximetry sensors. In this article a comparative analysis of various regression approaches - Support Vector Regression (SVR), Gradient Boosting, Lasso Regression, Ridge Regression and Sequential Neural Network-was carried out in order to discover an opportunity for predicting SpO2 level decreasing basing on the features of a spirogram for 4 different patients separately. The best results of R2 determination score were shown by models based on SVR and Neural Network: 0.6878 and 0.6697. The best correlation coefficient between the ground truth blood saturation decreasing and the predicted values was 0.9019 for the Support Vector Regressor model.