Detection of apnoea from respiratory time series data using clinically recognizable features and kNN classification.

Apnoea is a sleep related breathing disorder that is common in adults and can be described as a temporary closure in the upper airway during sleep. A system using time series analysis of one minute epochs of respiratory impedance signals to detect apnoea is described. An algorithm has been developed using MATLAB for extracting clinically recognizable features from the respiratory impedance signal. One minute samples are classified using kNN classification of the feature set. The output of the system has been shown to detect apnoeic episodes in eight eight-hour patient records collected from the PhysioNet database. The specificity of the classifier is 88.1% and the sensitivity is 95.7%. ROC analysis was performed and the area under the ROC curve is 0.9604. Future research will include testing the classifier in a much larger dataset and also a novel method for the presentation of classification results to physicians.