Lactate threshold testing accuracy

BSXinsight has been validated by many and performed live tests against blood testing frequently. One of the more recent validations can be found in the Journal of Strength and Conditioning Research.

BSXinsight is over 95% accurate at identifying lactate threshold. It uses novel algorithms and proprietary machine learning techniques built from literally hundreds (n=800+) of individual athlete tests to achieve this industry standard level of accuracy.

This means that across all athletes whom we’ve tested at BSX Laboratories we were able to identify lactate threshold using BSXinsight with 5% or less deviation from what independent professional assessment determined it to be using traditional blood sampling methods. In other words, those error measurements represent the percent deviation from what the blood tests determined lactate threshold to be using blood draws, and what BSXinsight determined it to be.

This approach is far superior to the basic regression techniques that have been published to date, because it eliminates the need for the identification of a 'break-point', which is especially susceptible to bias. In addition, our machine learning techniques are data driven, which allow the system to improve its predictive power over time, as more data is collected with each new athlete that trains with BSXinsight.

Our research team at BSX Laboratories is currently investigating how much of this error is due to true inaccuracies in BSXinsight’s data collection and/or technique vs. how much is due to the increased resolution that our approach has over traditional measurements and therefore represents an overall improvement in accuracy.

In other words, traditional blood sampling is taken every 2-3 minutes. BSXinsight on the other hand records thousands of data points every minute. This gives us much greater resolution and theoretically will allow BSXinsight to be more accurate than the old finger-stick method. We believe any discrepancy between the blood and needless method can eventually be explained by the limitations of the old method. These are things we're still investigating however.

Have more questions? Submit a request


Powered by Zendesk