The Detection of Abnormal Breathing Activity by Vision Analysis in Application to Diagnosis of Obstructive Sleep Apnea

Obstructive Sleep Apnea is increasingly seen as a common and important condition, contributing to sleep disturbance and consequential daytime sleepiness. This has potentially serious consequences for the individual, employers, and society as a whole. Apnea refers to short spells when breathing stops. In obstructive sleep apnea, the throat constricts during sleep, preventing breathing; the apnea episode often ends with a loud snore and/or gasp. Such an event is sufficient to open the throat muscles to allow breathing, and the patient usually falls asleep again so quickly that the event is not remembered. This cycle repeats itself throughout the night as the muscles relax and the throat blocks again, and the frequency of the episode is used to determine the severity of the syndrome. A wide range of parameters, including EEG (electroencephalogram) sleep staging, snoring, changes in airway resistance, airflow, and inspiratory effort, as well as oxygen saturation and body movement during “normal sleep” in a Sleep Lab, lead to an understanding of the pathophysiology of sleep apnea (Flemons et al, 2003). According to recent research findings (Bennett, Langford, Stradling & Davies, 1998; Kingshott et al., 2000; Pepperell, Davies & Stradling, 2002), the best predictors of morbidity in individual patients, as assessed by improvements with continuous positive airway pressure therapy (CPAP), are nocturnal oxygen saturation and movement during sleep. Moreover, Siven, Kornecki, and Schonfeld (1996) indicate that the results from traditional polysomnography are highly correlated with the video test results. Video monitoring has been adopted to assist diagnosis on obstructive sleep apnea. However, video monitoring and interpretation are less well developed, both from a recording and analysis viewpoint, due to the relative computational complexity of video analysis and the significant technical challenges. These include the lighting restriction involved with natural night vision and low illumination, including a lack of no color information; the obscuration of the body by a cover; the variation of human size, shape, and behavior; and the massive volume of video and audio data. The diagnosis of sleep disturbances requires a review of substantial amounts of data (including video footage) by clinicians. There is therefore a need for automated monitoring of human breathing activity to diagnose sleep disorders.