Project Description:
This projects is aiming to provide an affordable, accurate and seamlessly integrated method of detecting sleep apnea in an individual. We are using a classic machine learning model in order to detect if and when apnea occurs during a night's sleep. A single lead ECG signal, which can be collected and transmitted with wearables that are commonly found all over today's market, is used as the input which is then ran through our algorithm. We then receive the output, for each 1 minute segment of sleep, as 1 if apnea is detected or 0 if it is not detected. We were able to reach an accuracy of 82% using a K-nearest neighbors classifier. This algorithm has the potential of offering a more feasible alternative way to diagnose apnea to individuals who are unable to undergo a polysomnography test.
