Sleep Apnea sub-type detection from Polysomnography signals

One of the most common sleep-related disorder is Obstructive Sleep Apnea Syndrome (OSAS). Increased upper airway resistance during sleep causes partial or full airflow interruptions. In stroke patients, severe OSAS is increases the risk of mortality, neurological impairments, functional result after rehabilitation, and uncontrolled hypertension, making OSAS identification and treatment crucial. Polysomnography is the best OSAS test. In this work, it is proposed to use PSG signals to identify different sub-types of OSAS. Although many works have been done to detect OSAS but no work has been done till now to detect the sub-types of OSAS. It is of great medical importance to find out the sub-types of OSAS in order to plan proper treatment and diagnosis for the patient. Technically, it is more challenging to detect sub-types rather than just detecting the presence of OSAS. Deep learning architectures find use in extracting features from PSG signals. They further enhance learning to classify the 4 different categories of sleep apnea. Results demonstrate that a mean accuracy of 94.26% is achieved with the proposed methodology.