National projects

Title of the project Denoising Diffusion Probabilistic Models for Enhanced ECG Signal Noise Reduction in arrhythmia Classification, and AFIB Detection

Program Information technology for the development of science and the knowledge society

Project number Nr. S-ITP-25-9

Project implementation period 2025-10-01 – 2027-09-30

Head of Project dr. Jolita Bernatavičienė

Aim of the project – Cardiovascular diseases are a major cause of mortality worldwide, and accurate ECG analysis is important for early detection and pathological changes identification. However, real-world ECG signals are often corrupted by noise, such as baseline wander and electromyographic artifacts, reducing the reliability of arrhythmia and AFIB detection. This project aims to investigate and develop a novel approach for ECG signal noise reduction using Denoising Diffusion Probabilistic Models (DDPMs). By leveraging the strengths of DDPMs in modeling complex data distributions and generating high-quality samples, we aim to enhance the accuracy and reliability of arrhythmia classification and Atrial Fibrillation (AFIB) detection, particularly in noisy real-world ECG signals acquired from wearable devices like Zive'sECG monitor. The project will address the limitations of existing noise reduction techniques and contribute to advancements in personalized cardiac monitoring and AI-driven telemedicine. By reducing noise distortions, this project will significantly improve automated arrhythmia and AFIB detection, enabling earlier diagnoses and better patient outcomes. Additionally, synthetic ECG data generation using DDPMs will address class imbalances in medical datasets, enhancing AI model performance and reliability. This interdisciplinary collaboration between Vilnius University and Zive IAS leverages expertise in machine learning, biomedical signal processing, and real-world ECG monitoring. The project’s results will support telemedicine, wearable health technologies, and AI-driven cardiology, contributing to more accessible and reliable healthcare solutions.

The project is being carried out by a group of highly competent scientists. – dr. Jolita Bernatavičienė, dr. Jurgita Markevičiūtė, dr. Povilas Treigys, dr. Kęstutis Juškevičius.

Young scientists and students are also participating in the project and developing their skills.

Source of funding - Funding was provided by the Lithuanian Science Council (LMTLT), contract No. S-ITP-25-9.