Pushing the limits of spatial resolution for whole-heart cardiac MR angiography: Let’s get ready for prime time

This is a 3 year funded project for immediate start

Aim of the PhD Project:  

 In this proposal we aim to develop, implement and test the clinical feasibility of a novel deep learning based undersampled reconstruction to enable free-breathing submillimetre isotropic resolution 3D whole-heart coronary magnetic resonance angiography (CMRA) and multi-contrast cardiac magnetic resonance imaging in reduced scan times.  

 The specific objectives of the project are: 

1. To investigate a novel deep learning based undersampled reconstruction approach to accelerate the acquisition of 3D whole-heart CMRA. 
2. To enable free-breathing submillimetre isotropic resolution 3D whole-heart CMRA in clinically feasible scan times. 
3. To extend the proposed deep learning based reconstruction approach to accelerate the simultaneous acquisition of multi-contrast whole-heart coronary images (e.g. CMRA and black-blood plaque imaging).  
4. To validate the proposed approach in patients with coronary artery disease in comparison with current gold standards.  

Project description/background

Coronary Magnetic Resonance Angiography (CMRA) is a promising non-invasive tool for early risk assessment and monitoring of coronary artery disease (CAD), which avoids ionising radiation and nephrotoxic contrast agents in contrast to x-ray angiography and computed tomography. Multi-contrast coronary magnetic resonance imaging (MRI) provides the additional benefit of identifying high risk atherosclerotic plaques as it can provide unique information on plaque composition (e.g. intraplaque haemorrhage and thrombus) and plaque burden. However, CMRA is limited by image quality degradation due to respiratory motion, long and unpredictable acquisition times and lower anisotropic spatial resolution, compared to CTCA, which impedes the quantification of luminal stenosis and thus limits the utility of CMRA as a standalone investigation of CAD. To address this challenge, motion compensated isotropic sub-millimetre 3D CMRA in clinically feasible scan times is required for more accurate assessment of lesion severity and effective risk stratification of patients.

Recent developments from our group in the field of CMRA and multi-contrast whole-heart coronary imaging, including improved respiratory motion correction [1-2] and image acceleration with advanced reconstruction techniques [3] have enabled CMRA images in shorter and predictable acquisition times with high image quality. However, despite these developments, scan times for isotropic sub-millimetre resolution images remain lengthy. Furthermore, the long computational time required for this motion compensated undersampled reconstruction and the need for tuning of reconstruction hyper-parameters may hinder the application of these techniques in clinical practice. To address this challenge, here we propose to develop and test the clinical feasibility of a novel deep learning based undersampled reconstruction to enable submillimetre 3D cardiac MR images with in-line rapid reconstruction. The reconstruction should only take hundreds of milliseconds and thus could be incorporated in the scanner software to enable the use of these advanced techniques in clinical practice. In order to achieve this we propose:

• To investigate a novel deep learning based undersampled reconstruction approach to accelerate the acquisition of 3D whole-heart CMRA and thus enable free-breathing submillimetre isotropic resolution acquisitions in clinically feasible scan times.
• To extend the proposed deep learning based reconstruction approach to accelerate the simultaneous acquisition of multi-contrast whole-heart coronary images (e.g. CMRA and black-blood plaque imaging) assessing the potential for generalization and transfer learning of the proposed technique.
• To implement the proposed deep learning based reconstruction approach in the scanner software.
• To validate the proposed approach in patients with coronary artery disease in comparison with current gold standards.

This project encompasses topics from both the Emerging Imaging and AI in Medical Imaging themes of this CDT, and thus will permit the student to work at the interface of both sub-disciplines. The student will work closely with our industry partners who will provide support with the programming environment. Moreover, placements at Siemens Healthineers in Erlangen will be offered to jointly implement the solutions developed during this PhD in the Siemens scanning software.