Within this project we developed a set of deep-learning tools that enabled design of a robust, trustworthy, explainable, and transparent system, while retaining the superior level of performance expected of deep learning-based algorithms for classification of heart conditions from short ECG recordings collected using a two-lead device.
Within this project we developed an app which integrates an AI method that can automatically distinguish between atrial fibrillation, other rhythm disturbances and noise when using a mobile one-lead ECG device. In parallel we developed an automated AI-based system to identify heart conditions from 12-lead digital or image ECG recordings with high accuracy. We also demonstrated that the images scanned using a smartphone provided the same accuracy as machine images.
