Mining Contrast Rules in a Sleep Apnea Dataset

Sleep apnea is a common sleep disorder affecting a significant portion of the population around the world and is associated to severe health consequences. Existing sleep apnea diagnosis and screening methods require the use of expensive and invasive medical devices, which greatly limits the accessibility and affordability of those methods. The goal of our study is to develop a novel method for accurate sleep apnea screening with only a few physiological signals that can be readily measured with off-the-shelf wearable devices. At the current stage, we aim to develop an understanding into the types of signals and the specific statistical features that may help differentiate apnea and non-apnea episodes. To achieve the goal, we applied contrast set mining to an open sleep apnea dataset. The generated contrast rule suggests that a combination of higher heart rate, lower blood oxygen saturation, and a certain body movement pattern together was associated to the occurrence of sleep apnea events.