Abstract
Pulse wave analysis (PWA) is a technique that studies the shape of a pressure waveform caused by a heartbeat in order to estimate cardiovascular health. The waveform is often measured from the aorta, where the morphology is well understood in relation to cardiovascular physiology, but the waveform can also be measured from the peripheral circulation with a pulse oximeter, via photoplesmography. Thus, the goal of this project was to create a noninvasive diagnostic device that would measure a patient’s pulse waveform with the pulse oximeter, use an algorithm to transform it to the central aortic waveform, and use machine learning to analyze the shape of the waveform to screen for cardiovascular disease (CVD) risk. This project required three main goals; first, to perform a research study to collect the peripheral pulse waveforms of inpatients experiencing acute heart failure and outpatients in the catheterization lab. As a result, data was collected from 40 inpatients and 17 outpatients following Institutional Review Board (IRB) protocol. Second, to collaborate with the computer science department to develop an algorithm that would both transform the peripheral waveform to the central waveform, and analyze the key features in the waveform. As a result, a transfer function was generated using Fourier and Inverse Fourier transforms, using values validated in literature. Additionally, a thorough patient database including metrics of severity and outcomes for patients was created to use in machine learning development. The third aim was to develop an improved prototype with a user interface to facilitate future data collection, which would eventually become a finished product. As a result, a touch screen was added to the existing Arduino device, which has the capability of displaying raw data. Future work includes refining this prototype to process the raw data internally, and in-depth computer science techniques to analyze the data.
