Vascular Waveform Analysis Using Bayesian Pulse Deconvolution

Abstract: Vascular waveforms such as arterial pressure, photoplethysmography, and tonometry are widely used to measure vital signs, diagnose conditions, and predict long-term health outcomes. However, the analysis of these signals typically relies on heuristic filtering and signal processing operations that encode unexamined assumptions and introduce cascading errors and artifacts. Instead, we propose Bayesian pulse deconvolution as a flexible method for analyzing vascular waveforms informed by biological and physical domain knowledge. Across simulated waveforms and real human recordings we find this method significantly outperforms standard signal processing pipelines for filtering noise, detecting pulse times, and extracting pulse shapes.In this talk, I will explain the method and reflect on the advantages of analyzing human biosignals using Bayesian methods that encode domain knowledge and quantify uncertainty.

Bio: Parker Ruth is a Stanford Computer Science PhD student who researches health sensing systems and digital biomarkers for measuring cardiovascular and neuromuscular health. He is advised by Dr. James Landay, director of the Stanford Institute for Human-Centered AI, with close collaborators in biomechanics, bioengineering, statistics, cardiology, and radiology. He graduated summa cum laude from the University of Washington with bachelor's degrees in bioengineering and computer engineering. He has been named an NIH F31 Predoctoral Fellow, NSF Graduate Research Fellow, Hertz Fellowship Finalist, and Goldwater Scholar.