Monitoring and alerting system for sleep apnea patients

Abstract

Obstructive Sleep Apnea (OSA) is a very common sleep disorder which affects a significant portion of the population. The gold standard for monitoring OSA, and the treatment efficacy is through a Polysomnogram (PSG). A PSG measures a multitude of biological signals to determine the severity of OSA. However, a PSG is expensive, and causes discomfort for the patients. The system developed in the first phase of this project allows for home-based monitoring of OSA that is affordable and much more comfortable for the patient than the conventional PSG. It makes use of 3 biological signals, including a single channel frontal electroencephalogram (EEG) from the FP2-A1 positions, blood oxygen saturation (SpO2) reading, and recording of snores. In this system, every individual event is logged into a locally stored file, and if there are at least 2 OSA events happening at the same time, the OSA episode is deemed serious, and a buzzer will sound to wake the patient. A message is also sent to the caregiver / next-of-kin to notify them. The second part of the project focuses on a prediction algorithm, making use of a machine learning technique called k-nearest neighbours, to predict the onset of OSA. With such a prediction algorithm, it is possible to aid treatment techniques such as regulating airflow in the Continuous Positive Airway Pressure (CPAP) treatment.