Browsing by Author "Pereira, M. A. Pacheco"
Now showing 1 - 2 of 2
Results Per Page
Sort Options
- Predicting adherence to continuous positive airway pressure in patients with obstructive sleep apnea syndrome through machine learningPublication . Pereira, M. A. Pacheco; Teresa Ribeiro, RicardoMany patients with Obstructive Sleep Apnea Syndrome (OSAS) require Continuous Positive Airway Pressure (CPAP) therapy. Despite its high efficacy, both in the short and long term, treatment through CPAP has low adherence rates, even with the technological advances in recent years. In this study, using machine learning algorithms, we tried to predict which patients would be successful in adhering to CPAP treatment (mean =4h per night), three months after the beginning of the treatment, through the data obtained from a multicentre public database (n=175). After comparing six algorithms, Neural Networks (NN) was the one that showed the best results, with an f1-score of 0.71 and an AUC of 0.75, followed by Linear Regression, kNN, SVM, Naive Bayes, and Random Forests. Ten relevant characteristics were also identified for predicting adherence success: severity of OSAS, time til treatment, waist perimeter, score of FOSQ global, Apnea-Hypopnea Index, seizure diagnostic, type of sleep study (home vs. full night in laboratory vs. split night in laboratory), liver disease diagnostic and score FOSQ vigilance. It is possible to conclude that ML algorithms, properly trained in Big Data systems, may have a reasonable predictive capacity for the success of patients' adherence to CPAP, thus allowing personalized therapy with an improvement in their quality of life.
- VO2 prediction based on physiologic and mechanical exercise measurementsPublication . Pereira, M. A. Pacheco; Almeida, R.; Dias, Hermínia BritesThe Cardiopulmonary Exercise Test (CPET) is a diagnostic test that evaluates the functional capacity of an individual through the integrated response of the cardiovascular, respiratory, and metabolic systems. VO2 max is the parameter that access functional capacity, although it’s difficult to achieve given the effort that implies. In recent years, an increase in computing capabilities combined with the available storage of large amounts of information has led to a heightened interest in machine learning (ML). We aimed in this study to enable CPET with ML models that allow predicting oxygen consumption in healthy individuals. The study methodology is based on the cleaning and exploratory analysis of a public database with about 992 CPETs performed on healthy individuals and athletes. To predict each value of VO2 (~569,000 instances), five ML algorithms were used (Random Forests, kNN, Neural Networks, Linear Regression, and SVM) with heart rate, respiratory rate, time from the beginning of the exam and treadmill speed, using a 20-fold cross-validation. The best result came from the Random Forest model, with a R2 of 0.88 and a RMSE of 334.34 ml.min-1. Furthermore, using the same methodology but different features, we tried to predict the VO2max with the 724 adult participants with a maximal test (RER≥1.05) but weaker results were obtained (the best model was the Linear Regression, with a R2 of 0.50 and a RMSE of 498.06 ml.min-1). Still, this model showed a better correlation with the real VO2max than the Wasserman equation (R=0.71 vs R=0.59). It is possible to predict with accuracy breath-by-breath VO2, based on easy-to-obtain physiological and mechanical measurements.