Analysis sleep apnea using machine learning: One possibility

Obstructive Sleep Apnea (OSA) is a public health problem, sometimes under-reported due to the difficulty in its diagnosis. Therefore, health professionals seek alternative diagnostic methods based on clinical and epidemiological parameters. Here, was evaluated the predictive capacity of the anthropometric and clinical parameters of the free access database provided by Penzel et al7 and Dublin Sleep Apnea Database of the University Hospital / University College of St. Vincent in the diagnosis of OSA. The database is composed of 56 participants of both sexes, aged between 27 and 63 years. The following indicators were evaluated: weight, height, body mass index, sex and age. SPSS® software was used for statistical analysis of the database. Only low cost methods were used, reproducible and innocuous, reaching AUROC of 0.98, 0.96 and 0.94 for BMI, weight and age, respectively, in the prediction of OSA. The evaluated indices presented high power prediction from the OSA and, due to the reproducibility and ease of application, can be used in the construction of a classifier algorithm that will allow the early diagnosis of OSA.