Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Longitudinal Wrist PPG Analysis for Reliable Hypertension Risk Screening Using Deep Learning

Hui Lin, Jiyang Li, Ramy Hussein, Xin Sui, Xiaoyu Li, Guangpu Zhu, Aggelos K. Katsaggelos, Zijing Zeng, Yelei Li

TL;DR

This study leverages deep learning models, including ResNet and Transformer, to analyze wrist PPG data collected with a smartwatch for efficient hypertension risk screening, eliminating the need for handcrafted PPG features.

Abstract

Hypertension is a leading risk factor for cardiovascular diseases. Traditional blood pressure monitoring methods are cumbersome and inadequate for continuous tracking, prompting the development of PPG-based cuffless blood pressure monitoring wearables. This study leverages deep learning models, including ResNet and Transformer, to analyze wrist PPG data collected with a smartwatch for efficient hypertension risk screening, eliminating the need for handcrafted PPG features. Using the Home Blood Pressure Monitoring (HBPM) longitudinal dataset of 448 subjects and five-fold cross-validation, our model was trained on over 68k spot-check instances from 358 subjects and tested on real-world continuous recordings of 90 subjects. The compact ResNet model with 0.124M parameters performed significantly better than traditional machine learning methods, demonstrating its effectiveness in distinguishing between healthy and abnormal cases in real-world scenarios.

Longitudinal Wrist PPG Analysis for Reliable Hypertension Risk Screening Using Deep Learning

TL;DR

This study leverages deep learning models, including ResNet and Transformer, to analyze wrist PPG data collected with a smartwatch for efficient hypertension risk screening, eliminating the need for handcrafted PPG features.

Abstract

Hypertension is a leading risk factor for cardiovascular diseases. Traditional blood pressure monitoring methods are cumbersome and inadequate for continuous tracking, prompting the development of PPG-based cuffless blood pressure monitoring wearables. This study leverages deep learning models, including ResNet and Transformer, to analyze wrist PPG data collected with a smartwatch for efficient hypertension risk screening, eliminating the need for handcrafted PPG features. Using the Home Blood Pressure Monitoring (HBPM) longitudinal dataset of 448 subjects and five-fold cross-validation, our model was trained on over 68k spot-check instances from 358 subjects and tested on real-world continuous recordings of 90 subjects. The compact ResNet model with 0.124M parameters performed significantly better than traditional machine learning methods, demonstrating its effectiveness in distinguishing between healthy and abnormal cases in real-world scenarios.

Paper Structure

This paper contains 9 sections, 1 equation, 3 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Methods
  3. PPG Preprocessing
  4. Hypertension Detection
  5. Experiments
  6. HBPM Dataset
  7. Model Validation
  8. Results and Discussion
  9. Conclusion

Figures (3)

  • Figure 1: Overview of the end-to-end deep learning framework for hypertension detection. The framework consists of two key components: (1) PPG preprocessing, which includes detrending, low-quality beat removal, outlier removal, normalization, and segmentation, and (2) a ResNet-based feature learning and classification model that processes the preprocessed PPG data to predict hypertension risk, where Conv denotes convolutional layers, $N$ represents the number of residual blocks, and FC refers to fully connected layers.
  • Figure 2: Overview of the preprocessing pipeline for PPG signals. The raw PPG signal (top left) shows a baseline drift, which is removed in the detrended version (top right) where low-quality beats are identified in red. The bottom left panel illustrates outlier detection based on z-scores, with outliers marked in red outside the blue and green boundaries. The bottom right panel presents the signal after outlier removal, which is prepared for further processing.
  • Figure 3: Comparison of ROC and PR curves for three models. The red line is a reference. The plot shows that the ResNet-4 outperforms in both metrics.