How the Oxynet web applications are used to crowdsource and interpret cardiopulmonary exercising tests data

The cardiopulmonary exercise test (CPET) is the gold standard procedure to assess an individual's aerobic fitness. Oxynet is an ongoing research effort to deliver an objective and automatic interpretation of CPET. On its software side, Oxynet consists of a free-to-use crowdsourcing web-application and a free-to-use deep learning inference web-application. Particularly, Oxynet includes a ventilatory threshold detection deep learning algorithm. Oxynet can achieve expert-level performance in the identification of the first and second ventilatory thresholds from CPET data. Accuracy is preserved across individuals with different aerobic fitness. Declared average root mean square errors and worst 90th percentiles in terms of time and oxygen uptake for individuals with low/ moderate/high fitness levels are: 7/22/10 (90th = 13/38/18) and 12/17/8 (90th = 43/27/17) s, 22/43/92 (90th = 76/95/246) and 29/56/44 (90th = 69/106/109) mlO2/min, for the first and second ventilatory threshold, respectively. Oxynet can be considered the first example of a deep learning algorithm trained with crowdsourced data in the field of the cardiopulmonary exercising test and triggers new opportunities for collaboration between experts in the field of exercise physiology. This project can potentially provide low-cost and time-efficient universal access to cardiopulmonary exercise test interpretation.