A Machine Vision Enabled Implantable pH Sensing Scheme Using Ultrasonic Images of Microcomposite Hydrogels

Stimuli-responsive hydrogels with embedded microparticles (i.e., silica beads) can be implanted either subcutaneously or deeply to monitor the physiological signals under the epidermis in a remote manner. This can be achieved by using an ultrasonic imaging system to detect the gel deformation in response to the signal change. Unfortunately, it is practically difficult to detect the swelling dynamics of the gel accurately due to the misalignment of the gel and ultrasonic wave. To address this problem, a machine vision-based algorithm is developed. Its procedure is to process the recorded 3-D ultrasonic video using noise reduction, mathematical morphology, edge detection, polygonal approximation, and projection correction, in series. The gel’s irregular cross section is then converted to a square, allowing for the accurate determination of its cross-sectional area. In the end, a linear or logarithmic regression model is trained to couple the cross-sectional area of the gel to its surrounding pH value. The optimal result shows the highest ${R} ^{{2}}$ of 0.86 and the lowest RMSE of 0.97. This technique is expected to work with a variety of implantable, wireless hydrogel sensors featuring a portable ultrasonic imaging device.