Sensor Fusion for Lactate Estimation

Abstract

During intense physical extertion there is a accumulation of lactate in the body, and eventually the workload has to be reduced. This thesis presents a overview of the energy systems in the body connected to the production of lactate and methods for detecting the physiological parameters in athletic endevours. The most common way of detecting lactate per today is with taking blood samples and analyzing the result with hand-held devices. Because this is both expensive and not practical during periods of intense physical extertion, there is an interest to measure it non-invasively. The company Predictor Medical is developing a non-invasive sensor, the BioMKR for estimating Glucose levels in the body, and this thesis tries to examine if the unit can be used for estimating lactate as well. Experiments were done were test subjects exerted themselves to exhaustion, ensuring lactate accumulation while wearing several different sensors in addition to the BioMKR for measuring parameters such as heart rate, VO2, VCO2, SMO2, NIR, bioimpedance, while measurements of blood glucose and blood lactate was done for comparison. The thesis presents correlations from different sensor types and parameters. Multivariate analysis by partial least squares regression was done on the different data-sets. While the results suffered from corrupted and noisy data and in general was not satisfactory, it was possible to build models with somewhat accuracy for predicting lactate level for specific sets, but these failed in predicting on a general model verified by other data sets.