Measuring Exertion, Part 3

We talked about the problem of inaccuracy of activity trackers for managing CFS. In a nutshell, CFS patients are so sensitive to exertions, that the resolution and precision of trackers, which are intended to aid healthy people to get healthier, are not adequate enough to register the small differences that are material to CFS patients.

Furthermore, there is a problem of measuring the high-exertion (for CFS patients), short-duration events. There are enough oxygen in the bloodstream and fuel in the muscles to power the activity for a short duration. Therefore, the heart rate is not raised proportional to the exertion, and the tracker tends to smooth out the heart rate and calorie expenditure over the trailing time. In other words, there is a damping effect for short intense activities. This is why a few squats or burpees, which trigger post-exertional sickness for me, hardly register on the tracker while slow walking that doesn't cause post-exertional sickness racks up tons of calories and heart rates.

Yet another problem is that activity trackers are generally designed for motion-based aerobic exercise such as walking. (There is a reason why they are called glorified pedometers). They don't perform well tracking ADLs, which matters the most to CFS patients, or light strength exercises that CFS specialists often recommends.

To summarize, the problems with activity trackers are: 1) HR and calories are not accurate enough; 2) they can't track short, intense activities and 3) they can't track ADLs and strength exercises. To get around these problems, we need an activity tracker specialized for CFS patients.

One way is to detect the activity types relevant to CFS patients and then infer the amount of exertion from the type and the speed. This is no different than what CFS patients are already doing to pace themselves except that it will be automated and more consistent. Not only that, the accuracy can be improved through statistical and machine learning technique as more observations pile up.

Let's take walking as an example. We've seen that the calorie expenditure and heart rate measured by a tracker is not accurate enough to discern the difference of a few steps per minute. But by measuring the speed instead of the heart rate or calories, we can more accurately differentiate the exertion and predict the post-exertional sickness. Below is the plot of the speed for the walk on 7/03 that caused crash (red) and 6/24 that did not (black). We see the difference in the peak speed that caused the crash.

Similarly, if we can discern the activity of washing dishes, for another example, and measure the speed of it, we could deduce the amount of exertion from it.

Turns out, there are activity trackers that  do this for weight exercises and high intensity training. Theses trackers, however, are specialized for athletic training, and may not be suitable for CFS patients. They are tuned for truly high intensity activities of healthy athletes, and probably won't detect the intensities in CFS scale (which really means low intensity) and ADLs. But their existence means that the methodology is entirely possible and therefore can be adapted for activity tracking for CFS patients.

On the next installment, I'll talk about my progress with Android Wear device in implementing this. It'll take a while, so don't hold your breath yet.