R&D Project: AI-Powered Interpretation of Electrocardiograms
Developed an innovative tool for annotation and interpretation of electrocardiograms. Compatible with various kinds of ECG this product integrates artificial intelligence and deep learning to streamline cardiac patient care, monitoring and diagnostics.es cardiac patient monitoring and diagnostics.
is an off-the-shelf software for remote patient monitoring and clinical diagnostics to enable constant cardiac monitoring and patient health surveillance. The tool works with digital health devices and apps to interpret and annotate electrocardiograms.
Since the product’s main tasks are focused on healthcare, the main requirements were:
- HIPAA compliance and correspondence to the quality and performance standards for medical equipment
- Accuracy, providing near real-time accurate processing
- Reliability, working in remote and difficult locations
We developed an unprecedented tool to enhance and streamline the interpretation of electro cardio diagrams. Leveraging deep learning, the product detects and annotates a wide range of cardiac events according to HL7® aECG standard. Our highly adaptive solution can be integrated into an EHR system or work with a mobile health device. We developed a proprietary zero-knowledge system for the product to store Protected Health Information that is securely encrypted.
How It Works
The product is an all-encompassing system that integrates with health assets such as electrocardiographs, Holters, hospital EHR systems, etc. Our product is very flexible, enabling patient supervision in any location, from the regular checkup to cardiac monitoring during rehab.
- The electrocardiogram is obtained through a compatible digital recorder via API or a web-based platform and then automatically transferred to the system for interpretation. The team is constantly working to add new ECG formats to the list of supported.
- The data is then analyzed to recognize patterns and detect cardiac events.
- The software automatically calculates the amplitudes and intervals.
- The reporting is made in charts, diagrams, and tables, with marked points of interest.
Deep Learning and Artificial Intelligence
Pattern recognition lies at the core of The Product. Our team applied deep learning and big data approaches, analyzing the data from a ten-year population study. All records were labeled by the patent-pending technology and then verified by qualified cardiologists. These labels were used to train the system with the help of artificial neural networks.
As a result, we created a system that detects patterns so subtle that the doctors themselves can often miss, transforming patient monitoring and care.