3D Free-breathing MRI with High Scan Efficiency for Assessment of Cardiovascular Disease: Combining Acceleration and Motion Correction Techniques

Coronary artery disease (CAD) is the leading single cause of morbidity and mortality in the Western world, causing over 65.000 deaths every year in England. Improving the treatment and outcome of cardiovascular disease is one of the main priorities of the National Health Service in the UK. CAD reduces the blood supply to the cardiac muscle and can lead to chest pain (angina) or heart attack. Its diagnosis is currently performed by invasive, ionizing, and potentially harmful X-ray cardiac catheterization procedures. Magnetic Resonance Imaging (MRI) is a very promising non-invasive tool for early risk assessment, guidance of therapy and treatment monitoring of CAD. Clinical research studies have shown the potential of perfusion, cardiac function, late gadolinium enhancement (LGE; infarct) and coronary MR angiography (CMRA) for the assessment of CAD. However, technical developments are still needed to allow MRI achieving similar diagnostic accuracy as the current clinical gold standard. Two major challenges in cardiovascular MRI are 1) image quality degradation due to respiratory motion, and 2) long scan times resulting from the required high-spatial resolution. Both can lead to non-diagnostic image quality and a low patient throughput resulting in high operation costs if not adequately addressed. Current developments that facilitate these problems still suffer from either long acquisition times or require acquisitions under multiple breath-holds, which can be difficult in very sick patients and lead to a reduced diagnostic accuracy.

LGE and CMRA protocols are particularly affected by these limitations. LGE yields images with a high contrast between viable and fibrotic/infarcted myocardial tissue and requires high resolution to accurately measure infarct size and transmurality as it has important implications on treatment decisions and patient prognosis. Similarly high-resolution 3D acquisitions are required in CMRA to accurately visualize coronary stenosis. The most common approach to detect and correct for respiratory motion in free-breathing LGE and CMRA imaging is the use of diaphragmatic navigator echoes to gate the MR acquisition. With this approach the MR data is acquired only when the respiratory signal coincides with a predefined acceptance window of the breathing cycle (e.g. end-expiration) with all other data being rejected. Small gating windows of 3 to 5mm allow to reduce motion artifacts but lead to prolonged acquisition times, which may even result in scan abortion in patients with highly irregular breathing patterns. Current approaches often lead to long acquisition times and/or suboptimal image quality due to residual respiratory motion.

In this proposal we aim to develop, implement and test the clinical feasibility of an efficient (shorter scan time) and robust (neither planning nor patient collaboration required) respiratory motion compensation framework for LGE and CMRA. This framework should allow to reduce the acquisition time of current CMRA and LGE protocols from ~10 min to less than 3 min. We hypothesize that this can be achieved by developing a new 3D approach that combines novel data sampling schemes, novel undersampling reconstruction techniques and a mathematical framework for simultaneous undersampled image reconstruction and motion correction of the free-breathing acquired data.

In the near future this project will benefit the scientific research community working in the field of coronary artery imaging and myocardial tissue characterization, as well as those working on MRI reconstruction and motion correction. Moreover, it will strengthen the UK's already strong role in medical imaging research. In the long term, the beneficiaries of this research may be patients with coronary artery disease who will benefit from efficient and accurate non-invasive diagnostic tests. However, a broad clinical application would require subsequent clinical studies in a larger number of patients.

A simple approach to reduce respiratory-induced motion artefacts in multislice 2D cardiovascular MRI is to perform the acquisition under multiple breath holds (~15-20s each). However this approach may suffer from slice-misalignment due to different breath hold positions and requires patient cooperation, which can be challenging in severely ill patients. To overcome these problems 3D free-breathing navigator 'gated' acquisitions have been introduced to minimize motion artifacts by estimating the respiratory signal from the foot-head translational motion of the diaphragm. Gating is used to acquire MR data only when the respiratory signal coincides with a predefined acceptance window, preferably end-expiration, while all other data are being rejected. This approach has shown to considerably reduce motion artifacts when small gating windows are employed (3-5mm), however it leads to prolonged scan times since only a fraction of the acquired data is accepted for reconstruction (referred to as scan efficiency). Another drawback of the navigator-gated approach is the use of an oversimplified and patient-independent model for motion correction.

In this proposal we aim to develop, implement and test the clinical feasibility of an efficient (shorter scan time) and robust (neither planning nor patient collaboration required) respiratory motion compensation framework for late gadolinium enhancement (LGE) and whole-heart coronary MR angiography (CMRA). This method will achieve 100% scan efficiency by correcting all acquired data directly in the reconstruction after registration of high-resolution undersampled reconstructed 3D virtual image navigators. We hypothesize that this can be achieved by developing a new 3D approach that combines novel data sampling schemes, novel undersampling reconstruction techniques and a mathematical framework for simultaneous undersampled image reconstruction and motion correction of the free-breathing acquired data.

This research will benefit several groups. In the near future it will benefit the scientific research community working in the field of coronary artery imaging and myocardial tissue characterization, as well as those working on MRI reconstruction and motion correction. Moreover, it will strengthen the UK's already strong role in medical imaging research. In the long term, the beneficiaries of this research may be patients with coronary artery disease who will benefit from efficient and accurate non-invasive diagnostic tests. However, a broad clinical application would require subsequent clinical studies in a larger number of patients and to compare with the current clinical practice. In the long term this project will also help reducing the total scan time of CMRI examinations without affecting image and diagnostic quality. This may have two effects in the NHS: a) to reduce healthcare costs by decreasing the exam cost per patient, and b) to reduce the waiting time for MRI appointments by increasing the throughput of CMRI examinations. Through commercialization, the research will be of benefit to the imaging manufacturers and then to the healthcare system by improving patient management through improved treatment outcome and reduced healthcare costs on a national and international level.

Moreover this research will benefit the diagnosis and treatment of other cardiovascular diseases where whole-heart MRI and LGE acquisitions are required. An important example of such application is the use of MRI in X-ray guided interventions for treatment of arrhythmia (Knowles et al. 2010, IEEE Trans on Biomedical Engineering, 57(6): 1467-1475) or the planning of device implantation in patients requiring cardiac resynchronisation therapy (Duckett et al. 2011, JMRI, 33(1): 87-95). In these cases a high-resolution LGE MR image is needed to assess arrhythmic substrate or the location and extent of infarction whereas a whole-heart MRI is required as an overlay (roadmap) for image-guided intervention or for coronary vein visualisation.

The scientific methodology results from this research will be output as research publications in high-impact journals in the field of medical imaging. Target journals will include Magnetic Resonance in Medicine, Journal of Cardiovascular MR and Circulation. After the studies are published in scientific journals, and after they are patented if patenting is a viable option, data will be published on the study website for public access. We also plan to use the data to organize a "Motion Correction Challenge" in relevant Scientific Conferences. Dissemination will also take place through presentations at the major international conferences, especially the International Society for Magnetic Resonance in Medicine (ISMRM) and The Society for Cardiovascular Magnetic Resonance (SCMR). At the completion of this project we plan to share the developed scanner code (patches) with other MR researchers and will work closely together with Philips Healthcare to transfer the code into the product software to allow wide spread clinical use of this research. We will protect any resulting intellectual property in consultation with KCL Enterprises.