Speaker
Description
Cardiac Magnetic Resonance (CMR) is a non-invasive imaging modality and has become the gold standard technique for evaluating myocardial function, quantifying myocardial volumes and detecting myocardial scar. However, CMR requires highly trained operators with a good knowledge of cardiac anatomy and CMR exam planning. Due to the variability of cardiac morphology and body shape between patients, it can take considerable time for an operator to become proficient at planning the different cardiac planes. The examination times are also directly related to the training level of the operator; a proficient operator will save considerable scanner time. Thus, CMR is seen as a complex and time-consuming technique that is limited to high volume tertiary centres with operators specialised in CMR.
Recently, we have developed an automated deep learning-based planning tool [1] that performs the planning of the different cardiac planes without requiring user input. The goals were to 1) simplify CMR exams and reduce the overall training required for CMR operators; 2) provide reproducible and efficient studies with reduced examination time in all centres; 3) guarantee high-quality diagnostic images; 4) reduce the reliance on operators with advanced CMR training. Our current implementation provides the positioning for one specific view (e.g. small axis, two-chambers or four-chambers) but we believe that implementing an alternative architecture that would consider all three views at once would improve on the current performance.
Data required for this project are available through a collaboration with Linköping University (Sweden).
