Digital image correlation guided wearable sensor design and fabrication of novel microfluidic pump for human movement classification using machine learning models

Wearable technologies are transforming healthcare and enabling physicians to customize treatments through continuous non-invasive data acquisition. Microfluidic sensors have been gaining attention due to their flexibility and ability to conform to the human skin. However, individual variations in skin properties necessitate personalized sensor designs. This thesis presents a novel approach for designing and fabricating wearable microfluidic sensors, leveraging digital image correlation (DIC) to guide the development of a microfluidic pump for human movement classification. An open-source DIC platform Ncorr was utilized for 2D and 3D DIC analysis of human skin strain. The results were used to guide the design of a novel skin-strain-actuated microfluidic pump (SAMP) that employs asymmetric aspect ratio channels to record human activity in the fluidic domain. The passive image-based data readout method eliminates the need for electronics. An analytical model was established describing the SAMP’s operation mechanism as a wearable microfluidic device. A series of benchtop experiments established SAMP’s capability of measuring hundreds of strain cycles. Human experiments confirmed proof-of-concept by quantifying simple wrist flexion and distinguishing between complex shoulder movements. Finally, a theoretical method for improving the movement classification by utilizing machine learning models was discussed.