Photoplethysmogram (PPG)-Based Blood Pressure Estimation using Vision Transformer Networks: A Deep Learning Approach

Accurate blood pressure (BP) assessment is vital for effective cardiovascular health management. Traditionally, techniques such as Pulse Transit Time (PTT) utilized both Photoplethysmography (PPG) and Electrocardiography (ECG) signals for estimating BP. As the field progressed, there has been a notable shift towards estimating BP solely from PPG signals using Machine Learning (ML) techniques. Conventional ML models, while demonstrating notable achievements, encounter difficulties in fully capturing the complex temporal and spatial dependencies inherent in PPG data. Accordingly, deep learning methods, such as Convolutional Neural Networks (CNNs) and Long Short-Term Memory networks (LSTMs), are being explored by researchers to confront these challenges. Building upon this progress, this study aims to leverage the potential of Vision Transformers (ViTs) for BP estimation from PPG signals. This study optimizes a pre-trained Vision Transformer (ViT) model for accurate blood pressure (BP) estimation, utilizing Python for data preprocessing and Hugging Face libraries for fine-tuning. The research assesses the model’s performance on two distinct datasets, showcasing ViT’s adaptability and reliability in biomedical signal analysis. Furthermore, a comparative analysis is performed using a pre-trained ResNet50 model, which unveils comparable effectiveness with slight discrepancies. These results underscore the promise of ViTs and emphasize the importance of ongoing research to enhance model architectures and broaden datasets, aiming to achieve higher predictive accuracy in biomedical signal analysis.