Breath analyzer for personalized monitoring of exercise-induced metabolic fat burning

Obesity increases the risk of chronic diseases, such as type 2 diabetes mellitus, dyslipidemia, and cardiovascular diseases. Simple anthropometric measurements have time limitations in reflecting short-term weight and body fat changes. Thus, for detecting, losing or maintaining weight in short term, it is desirable to develop portable/ compact devices to monitor exercise-induced fat burn in real time. Exhaled breath acetone and blood-borne beta-hydroxybutyric acid (BOHB) are both correlated biomarkers of the metabolic fat burning process that takes place in the liver, predominantly post-exercise. Here, we have fabricated a compact breath analyzer for convenient, noninvasive and personalized estimation of fat burning in real time in a highly automated manner. The analyzer collects end-tidal breath in a standardized, user-friendly manner and it is equipped with an array of four low-power MEMS sensors for enhanced accuracy; this device presents a combination of required and desirable design features in modern portable/compact breath analyzers. We analyzed the exhaled breath (with our analyzer) and the blood samples (for BOHB) in 20 participants after exercise; we estimated the values of BOHB, as indication of the fat burn, resulting in Pearson coefficient r between the actual and predicted BOHB of 0.8. The estimation uses the responses from the sensor array in our analyzer and demographic and anthropometric information from the participants as inputs to a machine learning algorithm. The system and approach herein may help guide regular exercise for weight loss and its maintenance based on individuals' own metabolic changes.

Automated summary

Obesity increases the risk of chronic diseases, and it is important to develop portable devices to monitor exercise-induced fat burn in real time for weight management. In this study, a compact breath analyzer was developed for convenient and noninvasive estimation of fat burning, which showed a high level of accuracy using a machine learning algorithm.