Open ib00 opened 1 month ago
Thank you!
Out of curiosity, what type of sensor are you currently using?
Thanks for your reply and sharing this code of course.
I am a trail runner and a cyclist, so most of my efforts are longer and stochastic (due to terrain). I have Moxy which probably desaturates way too much. On certain types of intervals I can go from 80% to 20%. I also have Humon which has the opposite problem, not much dynamic range. I go from 75% to about 60% which is not much.
I haven't seen much use for hemoglobin concentration. Hopefully, you have some good ideas on how to put it into your model.
You're welcome.
Understood. This type of model actually performs quite well with unstructured workouts on mixed terrain.
The image above is from a ~4 hour bike ride at altitude and on mixed terrain (mountains). Ultimatley, I see these types of models being most useful in these contexts where traditional power-based intensity zones tend to fall short.
I haven't used this model on Moxy or Humon data since I currently do all of my data collection with the NNOXX wearable (which measures SmO2, NO, and acceleration). In my experience the range in SmO2 values provided by NNOXX is somewhere between what you'd see on a Moxy or Humon device, though NNOXX has more measurement-to-measurement variability since it's sampling at 50x/second whereas Moxy is 1x/ every other second (I don't know what Humon's sampling frequency is).
Like you, I don't see a lot of value in total hemoglobin (THb) measurements. Blood volume measurements, of that sort, are sensitive to many different factors such as mechanical effects of muscle contraction (compression, venous occlusion, etc), local muscle metabolic response (hypoxic vasodilation, changes in Hb's O2 dissociation curve, etc), and things like sympathetically mediated vasoconstriction. As a result, it tends to be a fairly 'dirty' signal, and often worsens these types of model's predictive values.
This is cool.
Few questions:
Thanks.