Sweat is a biomarker-rich fluid with potential for continuous patient monitoring via wearable devices. However, biomarker concentrations vary with the sweat rate per gland, posing a challenge for sweat sensing. To address this, we propose an algorithm to compute both the number of active sweat glands and their individual sweat rates. We developed models of sweat glands and a discrete sweat-sensing device to sense sweat volume. Our algorithm estimates the number of active glands by decomposing the signal into patterns generated by the individual sweat glands, allowing for the calculation of individual sweat rates. We assessed the algorithm's accuracy using synthetic datasets for varying physiological parameters (sweat rate and number of active sweat glands) and device layouts. The results show that device layout significantly affects accuracy, with error rates below 0.2% for low and medium sweat rates (below 0.2 nL min-1 per gland). However, the method is not suitable for high sweat rates. The suitable sweat rate range can be adapted to different needs through the choice of device. Based on our findings, we provide recommendations for optimal device layouts to improve accuracy in estimating active sweat glands. This is the first study to focus on estimating the sweat rate per gland, which essential for accurate biomarker concentration estimation and advancing sweat sensing towards clinical applications.
Read full abstract