Project 452581

Cardiomyopathy is a collection of conditions where abnormal health of the heart muscle can lead to heart failure, arrhythmias and death. It has become critically important to diagnose the cause of cardiomyopathy as tailored treatments are now available that can prevent progression or reverse some types of disease. However, these treatments must be started early and are reliant on correctly identifying the type of cardiomyopathy. Cardiac MRI is commonly used for this purpose, but is considered an expensive and complex test to perform and interpret. Our research laboratory has shown that using artificial intelligence (AI) we can extract information from the most simple and rapidly acquired MRI images, called "cine images". These beating heart images are routinely captured in around 5-8 minutes (compared to a full MRI scan which can take up to 1 hour and require additional use of contrast). Our team has developed an innovative approach to build a 3-dimensional "virtual beating heart model" from these simple images. These models can be interpreted by AI algorithms to detect if a cardiomyopathy is present and also identify what the underlying cause of disease is. If this novel approach is shown to work in other sites around the globe this may allow cardiomyopathy screening to be accomplished in less time, at lower cost, and with greater accuracy. We will use over 20,000 MRI scans collected by informed consent from patients in Alberta for the training of AI algorithms. Our algorithm will be trained to detect 12 different types of cardiomyopathy. To test if these algorithms will work in "real world" settings we will recruit another 2,500 patients from 11 other sites in countries from around the globe, each selected for their unique ethnicity profiles. We will test the performance of our AI algorithms according to both sex and different ethnicity to see if use of 3D cardiac modelling significantly improves diagnostic performance for under-represented groups.