Skip to main content
Slide 01

Deep Learning-Empowered ECG Perception for Heart Diagnostics

A collaborative research project for AI-enabled, automated and accurate arrhythmia detection and prediction.

Welcome to DEEP Heart Diagnostics, short for “Deep Learning-Empowered ECG Perception for Heart Diagnostics”, a cutting-edge research initiative at the intersection of artificial intelligence [AI] and cardiovascular diagnostics, hosted by the National and Kapodistrian University of Athens.

Here, we are pioneering advancements in medical technology to transform the landscape of heart diagnostics using deep learning [DL] and enable advanced and accurate diagnostic modules widely accessible and easily embeddable.

Concept and vision

Our concept revolves around harnessing the power of deep learning to revolutionize the way we perceive and diagnose heart conditions. The main goal is to compile a large amount of ECG data from various sources and apply sophisticated DL on them, to develop flexible and embeddable solutions for fast, accurate and explainable diagnosis of arrhythmias and heart conditions, with precision that exceeds that of human experts.

By fusing the expertise of cardiology with the capabilities of AI, we aim to unlock new dimensions in cardiovascular health. Our vision is to create a future where early detection of heart-related issues is seamless, accurate, and accessible to all. Through innovative technologies, we envision a world where medical professionals can make faster, more informed decisions, ultimately saving lives and enhancing patient outcomes.


  • Precision Diagnostics – Develop advanced algorithms for precise detection of arrhythmias and further cardiovascular abnormalities, beyond human precision.
  • Accessible Healthcare – Make state-of-the-art diagnostic tools easily accessible to healthcare providers worldwide.
  • Embeddable solution – Tune the developed solution so that it can be seamlessly integrated into existing monitoring devices and lightweight wearables.
  • XAI features – Apply explainable AI (XAI) technologies that allow clinicians to inteprete and trust model’s decisions, and also researchers to dive into and improve its structure.
  • Collaborative Innovation – Foster interdisciplinary collaboration between data scientists, medical professionals and researchers.


  • Early Intervention – Detect potential heart issues at an early stage, enabling timely and effective intervention.
  • Enhanced Accuracy – Provide healthcare professionals with accurate diagnostic insights, often surpassing human capabilities for certain tasks, aiming to improve decision-making.
  • Promotion of AI research in cardiovascular disease – Offer new insights on the benefit and applicability of various DL structures on biomedical data and promote related research.
  • Global Impact – Contribute to the global effort to reduce the burden of cardiovascular diseases.


The first step is to collect diverse and extensive sets of ECG data, ensuring a rich foundation for our subsequent analyses. Our approach emphasizes the acquisition of a wide spectrum of cardiovascular data, of various sources and modalities, that also entail several different disease entities. At the heart of our initiative lies the utilization of cutting-edge deep learning techniques. Our dedicated team opts for an iterated approach that employs sophisticated algorithms and neural network architectures to craft robust diagnostic models. Starting with novel convolution-based models, such as U-Net and 1D-convolution for time-series data, we proceed with applying a wide variety of DL-related hyperparameters, data preprocessing strategies and even additional architectures, recurrent networks of various structures.

Emphasizing innovation, we gradually scale up the training process with the use of heavy-duty, GPU-equipped servers and we push the boundaries of what is achievable in the realm of artificial intelligence, striving for excellence in model development.

Collaboration with domain experts is integral to our commitment to excellence. We subject our models to rigorous validation, leveraging the expertise of specialists in the field. This meticulous validation process, encompassing also the determination of the solution’s clinical benefit against human experts in the field, ensures the reliability and accuracy of our diagnostic models, meeting the highest standards in cardiovascular diagnostics.

Once validated, our solutions are strategically deployed. We make them accessible on platforms widely used by the medical and research communities. This deployment strategy aims to facilitate widespread adoption, ensuring that our validated solutions become valuable assets in the pursuit of advancing cardiovascular diagnostics. Through scientific events and publications, social media presence and further outreach activities we are committed to disseminate our research findings, along with the developed solution to the scientific community and the public.

Our emphasis on employing advanced deep learning techniques underscores our dedication to pushing the boundaries of innovation in the field. Through each phase of our workflow, we strive not only for scientific rigor but also for transformative advancements in the application of artificial intelligence to cardiovascular diagnostics.

Outreach and Dissemination

  • Global Scientific Events – Presenting our findings at global scientific events to engage with the medical and IT sectors.
  • Peer-Reviewed Journals – Publishing our research in reputable journals for wider dissemination.
  • Web Presence – Establishing a dedicated webpage and social media presence to provide updates and insights into our progress.
  • Outreach events – Hosting events for exposing our work and findings and engaging with the scientific community and the public.