Recent Advances in Cardiovascular Magnetic Resonance Techniques and Applications

Conventional CMR Reconstruction Techniques

Currently, most clinical CMR imaging is performed by sequentially acquiring lines of raw data on a Cartesian grid. This process is inherently slow, and without any acceleration techniques, it would take ≈250 ms to acquire a single image. A single-shot image with this temporal footprint would be corrupted by cardiac motion, and a cine acquisition would have only 4 frames per second. To overcome this limitation and increase effective temporal resolution, most CMR images are acquired in a segmented fashion, where a portion of the total required lines are acquired in each heartbeat during a breath hold, and the acquisition is synchronized to the ECG using cardiac gating. However, because images are synthesized from data acquired over multiple heartbeats, this technique is sensitive to changes in the cardiac cycle length, which occur during arrhythmias, and to changes in respiratory position because of inadequate breath holding. The situation becomes even more challenging for high-resolution, dynamic, or 3D acquisitions, which require even more data to be acquired. Thus, approaches for speeding up CMR image acquisition rely on using other prior information to produce images with fewer lines of data.

Parallel Imaging and k-t-Accelerated Imaging

The number of lines required to accurately reconstruct an image can be reduced using parallel imaging techniques which use differences in the spatial sensitivity of multiple receiver coils placed around the patient to provide the additional information needed for image reconstruction. Parallel imaging approaches are widely used on commercial magnetic resonance imaging systems with acronyms such as Sensitivity Encoding (SENSE), Generalized Autocalibrating Partial Parallel Acquisition (GRAPPA), and Autocalibrating Reconstruction for Cartesian imaging (ARC), but are typically limited to 2- to 3-fold acceleration factors because of the geometry of the coil elements and reduction in signal-to-noise ratio because of acquiring less data.

The amount of data needed for image reconstruction can be further reduced by taking advantage of the correlation between images at different times in the cardiac cycle or between different heartbeats in a dynamic acquisition. Images can be reconstructed with a reduced number of data lines using techniques such as k-t Broad-use Linear Acquisition Speed-up Technique (BLAST),¹ k-t SENSE, or k-t Principal Component Analysis (PCA).² These innovative techniques have enabled 3D perfusion imaging, 4-dimensional flow, and real-time imaging with high temporal and spatial resolution. Accelerations of 12× have been achieved using these techniques for 3D data acquisition.³ However, these techniques can be corrupted by respiratory motion.

Constrained Reconstruction and Compressed Sensing

An active area of image reconstruction research involves compressed sensing (CS), which enables reconstruction of images from significantly fewer lines of data by relying on the concept that the information content of CMR images is compressible, which is referred to as sparsity. CS has been used for several applications, including perfusion imaging,^4–6 flow imaging,⁷ angiography, T1 mapping,⁸ and real-time free-breathing cine imaging.⁹ The CS technique has also been applied to continuously acquired free-breathing radial techniques, such as in the Extra-Dimensional Golden-angle Radial Sparse Parallel Imaging (XD-GRASP)¹⁰ technique, which enables separation of the data in both cardiac and respiratory dimensions without the need for cardiac gating or breath holding (Figure 1) While these nonstandard reconstruction methods are often computationally demanding, the recent availability of improved computer hardware, such as graphical processing units, is making them practical. CS techniques are still in evolution and will need more research to determine whether they will be robust enough for routine clinical application.

Recent Technical Advances

T1 Mapping

Native T1 mapping is acquired before the administration of exogenous agents to provide a quantitative map of T1 relaxation times of the tissues being imaged on a pixel-by-pixel basis. Native T1 values reflect signals originating from the intracellular and extracellular (including interstitial and intravascular) compartments and reflect intrinsic differences in tissue properties. A range of T1 mapping techniques are available based on inversion recovery (eg, Modified Look-Locker Imaging [MOLLI],²⁶ shortened MOLLI [ShMOLLI]²⁷), saturation recovery (Saturation Recovery Single-Shot Acquisition [SASHA],²⁸ Saturation Method Using Adaptive Recovery Times for Cardiac T1 Mapping [SMARTT1]), or hybrid techniques (SaturationPulse Prepared Heart Rate Independent Inversion-Recovery [SAPPHIRE]).²⁹

Increased native T1 is useful for detecting acute myocardial pathologies, such as edema, infarction, and myocarditis,³⁰ and subacute cardiomyopathies, such as cardiac amyloidosis, hypertrophic cardiomyopathy, and dilated cardiomyopathy, and for assessing diffuse fibrosis.^29,31–34 Low global myocardial T1 values have clinical utility in myocardial iron overload (siderosis)³⁵ and Fabry disease.³⁶ In Anderson–Fabry disease, which is characterized by intramyocardial lipid deposition, low native T1 facilitates differentiation from other causes of left ventricular hypertrophy, such as hypertrophic cardiomyopathy and amyloidosis, which increase native T1. Native T1 mapping can detect the increase in myocardial blood volume associated with coronary vasodilation downstream of significant coronary stenoses.³⁷ This may potentially be an emerging application to detect myocardial ischemia without the need for GBCA. In addition to its ability to diagnose disease, there is increasing evidence that native T1 mapping carries prognostic power in risk stratification, including in patients with acute myocardial infarction, amyloidosis, and dilated cardiomyopathy.²⁹

T2 Mapping

Myocardial edema plays a role in many disease processes that affect the myocardium, including myocarditis and myocardial infarction. Prior to the widespread availability of T2 mapping pulse sequences, edema was typically imaged using dark-blood T2-weighted spin echo pulse sequences, but these sequences have several inherent limitations. Newly developed T2-mapping sequences can overcome these limitations by quantifying T2 relaxation times.³⁸ Studies in myocarditis and acute myocardial infarction have consistently demonstrated that T2 mapping provides increased diagnostic accuracy as compared with T2-weighted techniques for detection of myocardial edema.^39,40 T2-mapping techniques are replacing T2-weighted techniques for most diagnostic applications.

New Applications of T2* Mapping

T2* mapping has become well established for the detection of myocardial iron overload and predicts both the progression to heart failure and the requirement for future chelation therapy.⁴¹ However, there are several other important and emerging applications of T2* mapping.

Clinical Validation and Utility

Clinical validation using biopsies to correlate collagen content to ECV was first performed after a bolus of GBCAs followed by an infusion using a regular late gadolinium–enhanced sequence⁵⁰ and later also using ECV mapping techniques after a single contrast injection.⁵¹ The study by Ugander et al⁴⁸ nicely demonstrated the clinical utility of ECV mapping in patients with overt and subclinical myocardial pathology (Figure 5)

Several studies have demonstrated that T1 mapping and ECV mapping contribute mechanistic and prognostic insights into valve disease, infarction, and hypertension.^29,52 T1 and ECV have demonstrated potential utility in cases of suspected cardiomyopathy, infiltrative disease,⁵³ myocarditis,³⁰ and hypertrophic cardiomyopathy.⁵⁴ Widespread availability of fully automatic generation of ECV maps on the scanner will facilitate routine clinical use of this technique.⁵⁵

Challenges of Conventional Techniques

Motion of the heart resulting from both cardiac and respiratory cycles remains a challenge for robust CMR imaging. In patients with shortness of breath, breath holding may be inadequate, resulting in image artifacts. Free-breathing CMR techniques which use navigators to track diaphragmatic motion have been implemented in most MR systems from all major vendors and have been used in many CMR applications, such as noncontrast whole-heart coronary MRA,⁵⁶ 3D myocardial T1 mapping,⁵⁷ and 3D cardiac diffusion-weighted CMR.⁵⁸ However, relatively low gating efficiency, typically between 30% and 50%, may result in lengthy and unpredictable imaging times, particularly with whole-heart imaging. Recent navigator techniques which retrospectively correct motion, such as Three-Dimensional Retrospective Image-Based Motion Correction (TRIM), can achieve nearly 100% gating efficiency.⁵⁹ Although clinically useful, diaphragmatic navigator gating provides an indirect measure of heart motion and requires time and expertise to set up correctly. Several factors can cause poor ECG gating, particularly at 3T, resulting in motion-induced artifacts. While real-time imaging can be performed without ECG triggering, conventional techniques have reduced temporal or spatial resolution, limiting quality as compared with breath-held imaging.

Non-Breath-Hold and Non-ECG-Gated CMR Techniques

The CMR examination could be greatly simplified if high-quality images could be obtained without the need for breath holding or ECG gating, and this is an active area of CMR research. By significantly undersampling the raw data, it is possible to perform real-time CMR without ECG gating, which provides tolerance to arrhythmias. Non-Cartesian sampling patterns are often used for real-time cine imaging because they are more time-efficient and robust to flow and motion. With radial sampling and nonlinear inverse reconstruction with temporal regularization, Zhang et al⁶⁰ achieved a temporal resolution of 20 ms with online reconstruction using a graphical processing unit. Feng et al⁶¹ showed improved signal-to-noise ratio, contrast-to-noise ratio, and image quality of spiral acquisition over radial sampling for real-time cine CMR. Sharif et al⁵ used a continuous magnetization-driven radial sampling for non-ECG-gated myocardial perfusion magnetic resonance imaging. One of the problems with real-time CMR is limited temporal and spatial resolution. Kellman et al⁶² developed a retrospective reconstruction method to improve temporal and spatial resolution of real-time cine by combining data acquired over multiple heartbeats using respiratory motion correction and rebinning of data.

Direct measurement of heart motion can be obtained from a self-navigator derived from the imaging data itself.⁶³ Image projections,⁶⁴ 2-dimensional images,⁶⁵ or 3D⁶⁶ images have been used to derive self-navigator signals. In more recent techniques, motion data for both respiratory gating and cardiac triggering can be extracted by filtering a self-gating projection because cardiac and respiratory motion have different temporal frequency ranges and information content (Figure 6). These methods have been used for 2-dimensional⁶⁷ and 3D cine imaging⁶⁸ and 4-dimensional whole-heart coronary MRA.⁵⁹ An alternative approach is to resolve cardiac and respiratory motion using a CS technique, such as XD-GRASP, as described earlier.¹⁰ Although these techniques hold significant promise, they have yet to undergo significant clinical validation. These techniques may allow drastic simplification of the workflow or push-button CMR, which may facilitate the wide clinical application of CMR.