For this year-long medical device development project, I worked in a group of 4 to design and build a cuffless blood pressure monitor for early detection of preeclampsia in underserved communities. My role was to lead the neural network (U-Net) development to predict both systolic and diastolic blood pressure from photoplethysmography (PPG) signal data only. We were awarded with Columbia University's "Best Overall Design Award" in 2021. Below you can see our product executive summary, final hardware prototypes, overall design flow, final blood pressure prediction accuracies, a video of our data collection code running, a video of our prediction model being executed, and a simulation of our preeclampsia risk analysis app.
Collecting PPG data from a finger using Arduino and filtering the signal to create the correct deep learning model shape.
Running the U-Net in Python and reading the model's diastolic and systolic blood pressure predictions.
Prototype for supporting app.
