An adaptive CUSUM approach for automating sleep apnoea analysis based on pulse and oximetry data

Sleep apnoea is a common sleep disorder during human sleep. It is usually diagnosed by a doctor after recording one nights’ sleep signals. Patients have to go to the hospital to record sleep signals, which is time-consuming and resource-intensive. The study focused on two signals, pulse data and oximetry data, with the aim of detecting apnoea using a single signal. This paper introduced an anomaly detection approach using the adaptive cumulative sum (ACUSUM) change point detection algorithm to monitor outliers in the signal. In addition, the test results of ACUSUM will be compared with the test results of classical CUSUM. Besides, the threshold selection has been changed from an unchanging constant to a value related to the standard deviation of the selected signal based on a rational subgroup process. The results of the comparison confirm that ACUSUM is better than classical CUSUM in the accuracy of automatic detection.