Deep learning to predict power output from respiratory inductive plethysmography data

Abstract

Power output is one of the most accurate methods for measuring exercise intensity during outdoor endurance sports, since it records the actual effect of the work performed by the muscles over time. However, power meters are expensive and are limited to activity forms where it is possible to embed sensors in the propulsion system such as in cycling. We investigate using breathing to estimate power output during exercise, in order to create a portable method for tracking physical effort that is universally applicable in many activity forms. Breathing can be quantified through respiratory inductive plethysmography (RIP), which entails recording the movement of the rib cage and abdomen caused by breathing, and it enables us to have a portable, noninvasive device for measuring breathing. RIP signals, heart rate and power output were recorded during a N-of-1 study of a person performing a set of workouts on a stationary bike. The recorded data were used to build predictive models through deep learning algorithms. A convolutional neural network (CNN) trained on features derived from RIP signals and heart rate obtained a mean absolute percentage error (MAPE) of 0.20 (ie, 20% average error). The model showed promising capability of estimating correct power levels and reactivity to changes in power output, but the accuracy is significantly lower than that of cycling power meters.

Artificial Intelligence-Based Algorithms and Healthcare Applications of Respiratory Inductance Plethysmography: A Systematic Review

“…In [39], the authors found the limits of agreement of RIP with spirometry to be too high (≥20%) to validate RIP in obese people. The studies [36,41,42,52,[56][57][58][60][61][62][63]91] implemented machine learning and expressed their results in terms of accuracy, precision, sensitivity, specificity, and/or Cohen's kappa figures. The resultant figures were self-explanatory in terms of statistical significance.…”

“…Power output during exercise was studied in [60] with various predictive models, which showed moderate accuracy (R 2 = 0.56, mean absolute percentage error: 0.20-0.24) and need improvement. However, the fact that the data were obtained from a single subject is the greatest limitation of this study.…”

“…In [39,43], experiments were performed to determine the validity of RIP for high-BMI subjects. The studies [41,42,[60][61][62][63] solely focused on machine learning techniques for automated detection or prediction of non-ideal conditions. Studies on all applications of RIP identified in our search were included in this work having met the inclusion and exclusion criteria and passed the quality assessment.…”

Editorial

“…Helge has commissioned an article entitled ‘ Deep learning to predict power output from respiratory inductive plethysmography data ’ 1 . As we move into an age where our own personal health data are becoming increasingly important to personalise our lifestyles, including exercise regimes, there is a new scrutiny being placed on the equipment, which is required to generate these data.…”