Back to Projects

AI-enabled Imaging, Emerging Imaging

Advanced detection of myocardial ischaemia using cardiac MRI

Project ID: 2020_052

1st Supervisor and Clinical Champion: Amedeo Chiribiri, King’s College London
2nd Supervisor: Andy King, King’s College London
Additional supervisors: Teresa Correia and David Carmichael, King’s College London

Aim of the PhD Project:

  • Develop a safe high-resolution myocardial perfusion MRI protocol to obtain quantitative measurements of myocardial blood flow
  • Develop a deep learning approach to generate motion-corrected quantitative maps from highly undersampled data
  • Evaluate methods using simulated, dynamic phantom and patient data

Project Description / Background:

The aim of the project is to develop an efficient and safe cardiac magnetic resonance (CMR) technique to measure myocardial blood flow.

Coronary heart disease (CHD) is one of the leading causes of death in the UK. CHD is responsible for around 64,000 deaths in the UK each year. This disease occurs when coronary arteries become narrowed by a build-up of fatty substances within the walls (atherosclerosis), which can reduce the blood flow to the heart (myocardial ischaemia). A complete blockage of an artery can lead to a heart attack. Therefore, diagnosing myocardial ischaemia prior to a heart attack is important.

X-ray coronary angiography with fractional flow reserve (FFR) measurements is considered the gold standard for detecting CHD, but is invasive and uses ionising radiation [1]. Positron Emission Tomography (PET) is the clinical reference for non-invasive myocardial perfusion quantification [2]. However, PET is largely limited by availability and cost. CMR perfusion imaging is an emerging non-invasive imaging modality for detecting myocardial perfusion deficits [1,3,4]. It has advantages, such as higher spatial resolution, no radiation exposure, wider availability and lower scan cost compared to PET. However, CMR perfusion still suffers from several limitations, such as limited heart coverage, low spatial resolution, sensitivity to cardiac and respiratory motion (patients unable to hold their breath) and presence of dark-rim artefacts, which may be misinterpreted as ischaemia [3]. In addition, CMR perfusion imaging requires the use of potentially toxic contrast agents, particularly for patients suffering from kidney failure.

The aim of this project is to develop the next generation CMR perfusion technique that can be performed repeatedly without risk to the patient [4]. This will require the development and implementation of a novel arterial spin labelling (ASL) CMR acquisition sequence that overcomes current limitations while providing high-resolution quantitative CMR perfusion images. Moreover, to provide a more reliable diagnosis and risk stratification of patients, a method for accurately estimating the arterial input function will be developed, which is an essential component of tracer-kinetic model-based analysis [5,6].

This will also require the development of a deep learning-based reconstruction approach to generate motion-corrected quantitative maps [7,8] from highly accelerated free-breathing scans. This approach will be extended to automatically derive clinically relevant information from CMR perfusion scans. These new methods will be tested using simulations, our unique perfusion phantom [9] and in patients.

Candidates are expected to have a first-class honours degree in Physics, Computer Science or Biomedical Engineering or relevant Physical Sciences. Good knowledge of C++, Python or MATLAB is desirable. Experience with medical imaging, image reconstruction, pulse programming or machine learning is also desirable.

References:

[1] Nagel E, Greenwood JP, McCAnn GP, et al, Magnetic Resonance Perfusion or Fractional Flow Reserve in Coronary Disease. N Engl J Med. 2019;380(25):2418-2428
[2] Heo R, Nakazto R, Kalra D, Min J. Noninvasive imaging in coronary artery disease. Semin Nucl Med 2014;44(5):398-409.
[3] Fair M, Gatehouse P, Di Bella E, Firmin D.  A review of 3D first-pass, whole-heart myocardial perfusion cardiovascular magnetic resonance. J Cardiovasc Magn Reson. 2015; 17:68.
[4] Kober F, Jao T, Troalen R, Nayak K. Myocardial arterial spin labelling. J Cardiovasc Mag Reson. 2016; 18:22.
[5] Kellman P, Hansen M, Nieves-Vallespin S, et al. Myocardial perfusion cardiovascular magnetic resonance: optimized dual sequence and reconstruction for quantification. J Cardiovasc Magn Reson. 19(42), 1-14 (2017)
[6] Jerosh-Herold M. Quantification of myocardial perfusion by cardiovascular magnetic resonance.  J Cardiovasc Magn Reson. 2010; 12:57.
[7] Tourais J, Schneider T, Milidonis X, et al. High-Resolution motion-corrected 2D Myocardial Perfusion MRI using Locally Low-Rank and Wavelet Sparsity Constraints. ISMRM 2019, #1238.
[8] Henningsson M, Farias A, Villa A, et al. Quantitative myocardial perfusion using multi-echo Dixon for respiratory motion correction and arterial input function estimation. ISMRM 2018, #764.   
[9] Chiribiri, A., et al. Perfusion phantom: An efficient and reproducible method to simulate myocardial first-pass perfusion measurements with cardiovascular magnetic resonance. Magn Reson Med 2013; 69(3), 698-707.

Figure 1: Deep learning can be used to generate quantitative myocardial perfusion maps from undersampled cardiac magnetic resonance perfusion data. These maps can be used to identify perfusion deficits in patients with coronary heart disease

Back to Projects