Integrating classification and regression learners with bioimpedance methods for estimating weight status in infants and juveniles from the southern Cuba region.

Objective: The search for other indicators to assess the weight and nutritional status of individuals is important as it may provide more accurate information and assist in personalized medicine. This work is aimed to develop a machine learning predictions of weigh status derived from bioimpedance measurements and other physical parameters of healthy younger volunteers from Southern Cuba Region.

Methods: A pilot random study at the Pediatrics Hospital was conducted. The volunteers were selected between 2002 and 2008, ranging in age between 2 and 18 years old. In total, 776 female and male volunteers are studied. Along the age and sex in the cohort, volunteers with class I obesity, overweight, underweight and with normal weight are considered. The bioimpedance parameters are obtained by measuring standard tetrapolar whole-body configuration. The bioimpedance analyser is used, collecting fundamental bioelectrical and other parameters of interest. A classification model are performed, followed by a prediction of the body mass index.

Results: The results derived from the classification leaner reveal that the size, body density, phase angle, body mass index, fat-free mass, total body water volume according to Kotler, body surface area, extracellular water according to Kotler and sex largely govern the weight status of this population. In particular, the regression model shows that other bioparameters derived from impedance measurements can be associated with weight status estimation with high accuracy.

Conclusion: The classification and regression predictive models developed in this work are of the great importance to assist the diagnosis of weigh status with high accuracy. These models can be used for prompt weight status evaluation of younger individuals at the Pediatrics Hospital in Santiago de Cuba, Cuba.