Free-Running Cardiac and Respiratory Motion-Resolved Imaging: A Paradigm Shift for Managing Motion in Cardiac MRI?

Abstract

Cardiac magnetic resonance imaging (MRI) is widely used for non-invasive assessment of cardiac morphology, function, and tissue characteristics due to its exquisite soft-tissue contrast. However, it remains time-consuming and requires proficiency, making it costly and limiting its widespread use. Traditional cardiac MRI is inefficient as signal acquisition is often limited to specific cardiac phases and requires complex view planning, parameter adjustments, and management of both respiratory and cardiac motion. Recent efforts have aimed to make cardiac MRI more efficient and accessible. Among these innovations, the free-running framework enables 5D whole-heart imaging without the need for an electrocardiogram signal, respiratory breath-holding, or complex planning. It uses a fully self-gated approach to extract cardiac and respiratory signals directly from the acquired image data, allowing for more efficient coverage in time and space without the need for electrocardiogram gating, triggering, navigators, or breath-holds. This review provides a comprehensive overview of the free-running framework, detailing its history, concepts, recent improvements, and clinical applications.

Keywords: 5D; CMR; cardiac MRI; free-breathing; free-running; motion-resolved; self-gating.

Figure 1

One of the first MRI images of the heart was acquired in a volunteer using the Aberdeen NMR imager in 1979, before the spin-warp breakthrough. Although the contours of the body and some internal structures can be recognized, severe artefacts caused by cardiac motion were present preventing its clinical use. Image courtesy of Bill Edelstein.

Figure 2

Panel (A) Schematic overview of 3D radial sampling using a spiral phyllotaxis readout pattern (also known as an interleave). When the pre-defined number of radial spokes of a single interleave has been acquired, the next interleave will start which is rotated about the golden angle (~137.5°) from the previous one. Panel (B) With an increasing number of interleaves, an increasingly dense 3D kooshball of data points is obtained. After a pre-defined number of interleaves, the free-running acquisition is finished.

Figure 3

A schematic overview of enabling self-gating (SG) in 3D radial MRI by inserting a radial spoke in the superior-inferior direction at the start of each interleaf. Due to the continuous acquisition, an SG signal is obtained periodically as defined by the number of spokes per interleave (i.e., segments) and the repetition time (TR). With a TR of 3.5 ms and 20 segments per interleave (as shown), an SG signal is acquired every 70 ms and can be used to extract cardiac and respiratory motion information for total SG.

Figure 4

Panel (A) By extracting all superior-inferior (SI) spokes from each consecutive interleave, a self-gating (SG) signal is obtained. Panel (B) Using principal component analysis and filtering of this SG signal, the main respiratory and cardiac frequencies can be derived, and the respiratory and cardiac signature curves can be extracted. Panel (C) Using the respiratory and cardiac motion signals, all acquired radial spokes can be retrospectively binned into a variable number of cardiac phases and respiratory motion states. Using a compressed-sensing reconstruction, a 3D volume is obtained for each bin, ultimately leading to a 5D cardiac and respiratory motion-resolved dataset.

Figure 5

The obtained 5D image dataset can be visualised in various ways. (A) By fixing a specific respiratory motion state and looping through the cardiac phases, a cardiac 3D cine image is obtained (x-y-z-cardiac). (B) By fixing a specific cardiac phase and looping through the respiratory motion states, a respiratory 3D cine image is obtained (x-y-z-respiratory). (C) By fixing both a specific cardiac phase and respiratory motion state, a static 3D volume is obtained (x-y-z). For both dynamic 3D cines (A,B) and the static 3D volume (C), every desired (double) oblique cardiac view can be retrospectively selected using a multiplanar reconstruction.

Figure 6

Panel (A) Diastolic and systolic coronary reformats of paediatric patients after ferumoxytol injection under general anaesthesia and intubation, and under sedation and free-breathing. Data were collected using the 5D free-running framework. The orange arrowheads indicate the right coronary artery (RCA), while the blue arrowheads indicate the left main (LM) and left anterior descending (LAD) coronary arteries. It can be observed that image quality and vessel conspicuity appear similar in the two indicated cardiac phases, as well as that image quality is comparable between intubated and free-breathing subjects. Panel (B) Coronary reformats for three intubated and one free-breathing paediatric subjects for simultaneous visualisation of the RCA ostium (orange arrowheads) and LM artery ostium (blue arrowheads). The arrowheads with a red glow highlight anomalous coronary vessel anatomy. Figure created using the original source images by Roy et al. [29] with author permission.

Figure 7

Cardiac cine imaging in transversal, coronal, and sagittal views, and in left two-chamber, four-chamber, and short-axis views (using multiplanar reconstructions), acquired using the 5D free-running framework in a healthy adult subject using a 1-min and 6-min acquisition without contrast at 1.5 T. Although the significantly increased noise, lower image quality, and lower observer confidence for the 1-min acquisition are apparent, cardiac volumes and function were still comparable to reference standard 2D breath-hold cine imaging (not shown in this figure).

Figure 8

Panel (A) Magnitude images and phase-difference images for all three velocity-encoding directions (V_x, V_y, and V_z) in a transversal (TRA), coronal (COR), and sagittal (SAG) view for a specific time point during systole of a 41-year-old man with bicuspid aortic valve disease. Panel (B) Peak systolic velocity maximum intensity projections (MIPs) and streamlines show good agreement between conventional navigator-gated 4D flow and free-running 5D flow imaging, with some overestimation in the ascending aorta (AAo) and underestimation in the arch and descending aorta (DAo) as indicated by the white arrows. Panel (C) Flow curves at three locations in the aorta, as indicated by the red lines in panel (B), demonstrate good agreement between the two techniques. The underestimation in the arch and DAo for 5D flow can be observed. LV = left ventricle, SVC = superior vena cava. Figure created using the original source images by Ma et al. [52] with author permission.

Figure 9

Panel (A) Water-only and fat-only images, and corresponding parametric maps of water fraction and fat fraction of a healthy adult subject as obtained using the proposed 6D free-running framework (x-y-z-cardiac-respiratory-echo) at 1.5 T. Panel (B) Cardiac fat fraction maps in a transversal view for each phase of the cardiac cycle during end-expiration as obtained using the proposed 6D free-running framework in a healthy adult subject at 1.5 T, allowing tracking of the displacement of pericardial fatty regions throughout the cardiac cycle. Figure created using the original source images by Mackowiak et al. [58] with author permission.

Figure 10

Panel (A) Comparison of anatomical images acquired by the 5D free-running framework using a standard balanced steady-state free-precession (bSSFP) and fast-interrupted steady-state (FISS) acquisition at 3 T. The 5D datasets were multiplanar reformatted to a sagittal view, and images are shown in systole and diastole at both end-expiration and end-inspiration. The yellow lines help to indicate respiratory motion. The red arrows indicate the decrease in streaking artefacts in FISS compared to bSSFP. Panel (B) Comparison of coronary reformats obtained using both bSSFP and FISS, at both 1.5 T and 3 T. The water–fat cancellation artefacts at the coronary vessel borders can be observed in the bSSFP images, while these are absent in the FISS images. Panel (C) Comparison of anatomical images using a bSSFP and FISS acquisition at 1.5 T. Both 5D image data were binned into 16 cardiac phases and four respiratory motion states, of which eight cardiac phases in the end-expiratory motion state are shown in coronal view for each method. The red arrows indicate cardiac regions containing fat that is suppressed using FISS but not using bSSFP. Panel (D) Comparison of anatomical images using a bSSFP and FISS acquisition at 3 T. Both 5D image data were binned into 26 cardiac phases and four respiratory motion states, of which eight cardiac phases in the end-expiratory motion state are shown in coronal view for each method. The red arrows indicate cardiac regions containing fat that is suppressed using FISS but not using bSSFP. Figure created using the original source images by Bastiaansen et al. [69] with author permission.

Figure 11

Cardiac cine imaging in various short-axis, and left two-chamber and four-chamber cardiac views (using multiplanar reconstructions), acquired using the 5D free-running framework in two healthy adult subjects using a 7 min acquisition without contrast at 0.55 T. Despite the significantly increased noise and decreased blood–myocardium contrast at low field compared to 1.5 T, cardiac anatomy and function can still be depicted in detail using the fully automated free-running framework without the need for an ECG signal or repetitive breath-holding.