How Machine Learning Can Help Improve Heart Rate Measurements
Thanks to new ongoing research, heart rate monitors and fitness trackers may become a lot more useful in the future, which will have excellent implications for monitoring more serious health issues.
Heart Rate monitors have come a long way in recent years, and thanks to smart watches and other popular consumer devices, they’re now common among fitness enthusiasts.
They’ve very much gone from a “nice to have” to “I can’t believe you don’t have one”. Sales of 97.6 million in 2015 have risen to 130 million in 2018.
While most people use them to improve their time and track progress, the applications of heart rate monitors can be much more important.
How Monitors Can Have A Bigger Impact
Heart Rate Variability (HRV) can be used as a predictor of both psychophysiological stress and heart disease. Monitoring HRV in everyday life with smartwatches can make a huge difference in both predicting, and diagnosing these issues. With more and more people using less invasive fitness trackers and smart watches, the potential wider application is huge.
The problem though, lies in how accurate (or inaccurate) these devices can be.
While the technology has come along way, the fundamental issue with the accuracy of heart-rate monitors comes from the stability of the sensor. A recent study in Switzerland found a statistically significant correlation between error in HRV measurement and wearer movement. It should be noted that professional heart rate monitors, the types that use nodes attached to the chest, are still accurate. It is only the consumer brands that have accuracy issues.
What’s worse, the more you move, the more error prone these devices can be.
Fairly Well Documented Inaccuracies
Earlier this year, many news outlets highlighted how fitness trackers can be overestimating calorie burn, or inaccurately tracking mileage. In previous years, people have performed experiments wearing multiple devices at once and noticed a large variance among their readings.
The good news is that this same study has also devised a way to compensate for this, through machine learning:
We show that this error can be minimized by bringing into context additional available sensor information, such as accelerometer data. This work demonstrates our research-in-progress on how neural learning can minimize the error of such smartwatch HRV measurements.
More Serious Applications Of Heart Rate Monitors
While most people use heart rate monitors to improve their time or track their progress, heart rate monitoring can have much stronger benefits as well. As mentioned above, HRV has been shown to have a positive correlation with hypoglycemia, stress, and other heart diseases. The better we can become at monitoring HRV, the better we will be equipped to tackle these serious issues.
Currently, smart watches gather their data in a number of non-invasive ways, as explained in the mentioned report:
For instance, there are wrist-based consumer smartwatches that provide multiple data dimensions such as inter-beat intervals (obtained via optical sensor), three-axis accelerometer data, steps, burned calories or proprietary stress values. However, such non-professional devices need to be compared to a more precise instrument.
The Firstbeat Bodyguard 2 (heart rate monitor) is an inter-beat interval recorder employing two electrodes on the chest for measurement and that can be considered as a semi-professional device . Smartwatches thus collect a magnitude of data, most of which is sufficient for giving their user an overview of their daily activity. However, when being applied to more serious medical and healthcare use cases, the current measurement accuracy of wearable devices remains rather unclear.
This lack of accuracy makes smart watches somewhat limited in their ability to do serious help.
How Predictive Modeling Can Tackle This Problem
The researchers found, through ongoing research, that predictive modeling can help to significantly offset this inaccuracy. In short, by looking at how much variance occurs between everyday smart watches and more professional heart rate monitors, it’s possible to systematically predict how inaccurate a smart watch is, and then compensate for that to produce significantly more accurate data.
We found strong evidence that systematic errors in HRV measurements from the consumer smartwatch can be minimized with the utilization of additional data by the use of neural networks.
In the chart above, we can see that the accuracy of an adjusted fitness tracker (green line) is significantly better than that of one before adjustment (red line). The black line is the reference line, or the accuracy of the professional heart rate monitor (the most accurate form monitor).
This is very promising research and as technology improves, we should see the inaccuracy of consumer smartwatches minimized even more.
Not only will this mean more breakthroughs in serious health problems, but you’ll get more accurate data tracking with your casual jogging sessions too.
Read the full research paper here: https://arxiv.org/pdf/1907.07496.pdf