UC Irvine

As the next step in mobile health monitoring, an embedded system for monitoring the health of pregnant women and their fetuses in the home setting is nearing the brink of possibility. The main obstacles currently in the way are heavy algorithms and real-time performance, which have been solved with lighter computation for fetal heart rate on the system’s hardware (rather than via the cloud and through a server). This thesis will focus on the research done in optimizing the quality of electrocardiogram (ECG) sensor data so that the accompanying algorithm can return the most accurate fetal heart rate possible, which includes providing a powerful microcontroller for the algorithm to compute on and incorporating photoplethysmogram (PPG) data to help identify ECG peaks (which are vital to finding heart rate) that may otherwise have not been detected. Components for this system were selected to have a balanced tradeoff between sensor accuracy (to be able to capture ECG and PPG at reduced voltages due to abdominal placement) and power conservation (to minimally utilize the PPG and reduce energy during wireless transmissions). As part of ensuring proper connection between the components of the fetal system, outputted ECG simulator data from two different datasets and sensor-retrieved ECG data were compared using cross-correlation and showed 98% similarity between the two signals. Timing measurements for sensor-reading and algorithm functions were then taken by toggling a GPIO pin on the microcontroller and using an oscilloscope to read how long the pulse stayed high. These results caused several small modifications to the system in order to compensate for longer-than-expected function calls, with the most significant change being a decrease in the amount of data that can be processed by the algorithm. Simulator data being run through the algorithm confirmed a corresponding loss in accurate heart rate calculations (due to less detected peaks), but adequate accuracy in pinpointing the time indices of the signal’s peaks. Simultaneous sensor readings were taken to finally examine how well PPG data would be able to help calculate heart rate in the event of unidentified or misidentified ECG peaks, and results showed impressive correlation. Should these PPG experiments be further investigated and the algorithm optimized to process more data as intended, this monitoring system can make great progress towards being able to provide accurate fetal data more consistently throughout the course of pregnancies.