Estimating Vascular Stiffness through Speckle Plethysmography Waveform Analysis

Ali Owji

Mentor: Dr. Helen Parker, National Institute of Child Development (NICHD).

Date/Time: August 22nd, 2024 at 4:20pm.

Abstract: Noninvasive wearable devices for measuring biometrics such as blood pressure and vascular stiffness represent the future of personalized medicine and health monitoring. While pulse oximeters using photoplethysmography (PPG) are widely used to measure oxygen levels in the blood, Speckle Plethysmography (SPG) offers a promising alternative (Ghijsen et al., 2018). SPG uses an imaging technique to capture blood flow variations, producing more information-rich waveforms than PPG, with more defined peaks and troughs, including the systolic peak, diastolic peak, and dicrotic notch.

These waveforms can be used to estimate pulse wave velocity (PWV), a key indicator of arterial stiffness. Arterial stiffness is an important measure of cardiovascular health, with higher PWV indicating increased stiffness and potential health risks. This study explored SPG waveform features to correlate with PWV, aiming to develop a predictive model for estimating arterial stiffness. The study cohort included patients with CADASIL, a rare inherited disease affecting cerebral blood flow, as well as healthy controls.

A comprehensive derivative analysis was performed on the SPG signals, with both first and second derivatives calculated for each heartbeat. From these, 110 features were extracted, including timespans, ratios, amplitudes, and slopes. The analysis was conducted in two phases: an initial study with 9 patients, followed by an expanded study with additional patients and refined methodology.

The initial analysis, using z-score normalization, showed poor feature correlations with PWV, with the strongest correlation (r = -0.673) found in the amplitude ratio between the Diastolic and Systolic peaks. The follow-up study, incorporating lessons from PPG research, yielded results more consistent with previous findings, highlighting the age index (r = .775) and b/a ratio in the second derivative signal (r = .678) as strong predictive features (Hellqvist et al., 2024).

A Principal Component Analysis (PCA) revealed a nonlinear relationship between the features and PWV. This suggests that further research, including larger patient datasets, is necessary before developing a robust machine learning model for predicting arterial stiffness from SPG waveforms.

This study lays the groundwork for developing noninvasive, wearable devices capable of continuous vascular health monitoring, potentially revolutionizing preventive healthcare and personalized medicine.