Blood Pressure Prediction Using Ensemble Rules during Isometric Sustained Weight Test.

BACKGROUND: Predicting beat-to-beat blood pressure has several clinical applications. While most machine learning models focus on accuracy, it is necessary to build models that explain the relationships of hemodynamical parameters with blood pressure without sacrificing accuracy, especially during exercise. OBJECTIVE: The aim of this study is to use the RuleFit model to measure the importance, interactions, and relationships among several parameters extracted from photoplethysmography (PPG) and electrocardiography (ECG) signals during a dynamic weight-bearing test (WBT) and to assess the accuracy and interpretability of the model results. METHODS: RuleFit was applied to hemodynamical ECG and PPG parameters during rest and WBT in six healthy young subjects. The WBT involves holding a 500 g weight in the left hand for 2 min. Blood pressure is taken in the opposite arm before and during exercise thereof. RESULTS: The root mean square error of the model residuals was 4.72 and 2.68 mmHg for systolic blood pressure and diastolic blood pressure, respectively, during rest and 4.59 and 4.01 mmHg, respectively, during the WBT. Furthermore, the blood pressure measurements appeared to be nonlinear, and interaction effects were observed. Moreover, blood pressure predictions based on PPG parameters showed a strong correlation with individual characteristics and responses to exercise. CONCLUSION: The RuleFit model is an excellent tool to study interactions among variables for predicting blood pressure. Compared to other models, the RuleFit model showed superior performance. RuleFit can be used for predicting and interpreting relationships among predictors extracted from PPG and ECG signals.

Cardiac vagal imbalance to the isometric sustained weight test in adolescents with emotional eating behavior.

PURPOSE: To assess the relationship between emotional eating behavior and heart rate variability in Spanish adolescents during an isometric exercise test. METHODS: Participants included 52 adolescents aged between 13 and 18 years old. Heart rate was continuously recorded at rest (2 minutes) and during the sustained weight test (2 minutes). Linear and nonlinear methods of heart rate variability were assessed and related to the emotional eating behavior divided in two clusters. RESULTS: Statistically significant differences were observed in linear and non-linear parameters of heart rate variability comparing rest and sustained weight test. An increase in the value of emotional eating in overweight adolescents was founded. During the sustained weight test, there were differences between the two emotional eating clusters regarding the variables peak high frequency power, normalized low frequency power, normalized high frequency power, low frequency/high frequency ratio, and sample entropy. A positive correlation between the emotional eating behavior and the peak high frequency power was observed, though the prediction capacity of the high frequency waves is low it is observed that there is a good fit to the regression line. CONCLUSION: Results of this study shows that there was a relationship between vagal tone and emotional eating behavior in adolescents during an isometric exercise, with excessive parasympathetic predominance and sympathetic withdrawal during a physical effort.