Development of a software tool for automatic arousal scoring with application to polysomnographic analysis

Arousals frequently appear in subjects with sleep disorders (obstructive sleep apnea, OSA, for instance). Therefore, scoring and detecting arousals is a key point in order to assess the quality of sleep. Typically, this scoring task is manually carried out by an expert, according to his/her expertise. In this coordinated proposal, the GTAS research group collaborated with a group of physicians with a broad experience on sleep disorders. The aim of the project was to develop a software tool for automatic arousal detection and scoring. To this end, polysomnographic records obtained in a number of subjects (adults) with OSA were independently staged by several (2 / 3) experts who also marked the presence of arousals in every segment. These labeled data were used to develop parametric (Bayesian, for example) as well as non-parametric (for instance, Neural Networks or Support Vector Machines) detection algorithms. The results provided by the proposed detectors were validated by blind tests. The detectors were programmed in Matlab and integrated in a software tool that runs in a PC.