The increasing prevalence of cardiovascular diseases (CVDs) highlights the need for affordable and easily accessible tools for continuous cardiac monitoring. While Electrocardiography (ECG) is considered the standard, continuous monitoring remains a challenge. As a result, there has been interest in exploring Photoplethysmography (PPG) as a more basic alternative available in consumer wearables. This has led to efforts to translate PPG signals into ECG signals. In this study, we present the Region-Disentangled Diffusion Model (RDDM), a new diffusion model designed to capture the complex temporal dynamics of ECG. Traditional diffusion models like Denoising Diffusion Probabilistic Models (DDPM) struggle to capture these nuances due to the indiscriminate addition of noise across the entire signal. RDDM overcomes this limitation by selectively adding noise to specific regions of interest (ROI) in ECG signals, such as the QRS complex, and then disentangling the denoising of ROI and non-ROI regions through a reverse process. Our quantitative experiments demonstrate that RDDM can generate high-quality ECG signals from PPG in as few as 10 diffusion steps, making it highly effective and computationally efficient. To further validate the usefulness of the generated ECG signals, we introduce CardioBench, a comprehensive evaluation benchmark for various cardiac-related tasks, including heart rate and blood pressure estimation, stress classification, and the detection of atrial fibrillation and diabetes. Our extensive experiments show that RDDM achieves state-of-the-art performance on CardioBench. To the best of our knowledge, RDDM is the first diffusion model for cross-modal signal-to-signal translation in the bio-signal domain.
A Diffusion Model for Accurate PPG-to-ECG Translation with Region Disentanglement.
