Developing an Engine for Heart Failure Detection Using Deep Learning
HeartKinetics’ engine for heart failure detection has been improved and optimized. In this 8-week challenge, a collaborative team of 50 AI engineers from all around the world worked on the project.
Heart failures affect more than 60 million people worldwide. The main reasons are late diagnosis and a lack of resources to ensure proper follow-up. Thanks to the surge in e-health, this is set to revolutionize cardiology in the coming years.
The solution is based on artificial intelligence. The main objective of this project would be to apply Deep Learning to a set of records containing transformed cardiac signals (time series) to detect heart failure.
The project outcomes
The main goal of this project has been to ensure that for each input record, the model has predicted the ‘probability of Heart Failure’.
The databases that have been used during the project include:
Transformed signals, derived from the HeartKinetics proprietary database, which has contained over 500 records from both healthy subjects and subjects with heart failure.
The Frontiers public database consists of 100 records from healthy subjects and subjects with other cardiovascular diseases.
The IEEE Dataport public dataset, comprising 29 records from healthy subjects.