Overview
This project explores non-invasive methods for assessing cardiovascular health through wearable technology. Using speckle plethysmography waveforms from patients in the NIH Natural History Study of CADASIL, I developed analytical methods to estimate vascular stiffness without requiring specialized medical equipment.
The Problem
The project addresses a critical need in cardiovascular monitoring: traditional pulse wave velocity measurements are invasive and require clinical settings, limiting their use for continuous health tracking and early disease detection. There is a growing need for accessible, wearable solutions that can provide continuous cardiovascular monitoring.
My Approach
I implemented a comprehensive signal processing pipeline including:
- Automated Heartbeat Segmentation: Developed algorithm to identify and extract individual heartbeats from continuous waveforms
- Feature Engineering: Extracted 95 distinct features from SPG waveforms and their derivatives
- Statistical Analysis: Applied linear regression to identify strongest predictors of pulse wave velocity
- Physiological Validation: Confirmed findings across both healthy individuals and CADASIL patients
Key Findings
Through linear regression analysis, I identified the diastolic-to-systolic amplitude ratio as the strongest predictor of pulse wave velocity (r = -0.673, R² = 0.453). The findings reveal that healthier, more elastic arteries show higher diastolic-to-systolic ratios, while stiffer arteries demonstrate lower ratios.
This relationship holds across both healthy individuals and CADASIL patients, demonstrating how wearable SPG devices could provide accessible cardiovascular monitoring and early detection of vascular dysfunction in at-risk populations.
Technical Details
Features extracted included:
- Time spans between cardiac landmarks
- Slope measurements at critical points
- Area under curve calculations
- Amplitude ratios between systolic and diastolic peaks
- First and second derivative features
Technical Stack
Research Poster
Click the poster to view it fullscreen.
