Cardiac MRI, Technical Aspects Primer

Updated: May 16, 2022
Author: Eugene C Lin, MD; Chief Editor: Eugene C Lin, MD 

Overview

Starting from an almost experimental level in the 1990s, magnetic resonance imaging (MRI) of the heart has now become a routine tool within cardiac diagnostics. Besides its superiority in determining mass and volumes of the heart, structural imaging of myocardial tissue has become the domain of cardiac MRI. This implies detection and quantification of scars and fibrosis, identification of inflammatory and infiltrative processes in myocarditis and amyloidosis, and characterization of cardiac tumors. Perfusion imaging allows detection of ischemia and measurement of blood flow quantifies cardiac shunts and valve disorders. In combination with positron emission tomography (PET), the relationships of molecular/cellular processes and functional microstructural alterations become visible in myocarditis and amyloidosis through MRI.[1]

Cardiac MRI has a wide range of clinical applications. Many of these are commonly employed in clinical practice—for example, evaluation of congenital heart disease, cardiac masses, the pericardium, right ventricular dysplasia, and hibernating myocardium.[2, 3]

Other applications, such as evaluation of myocardial perfusion and of valvular and ventricular function, are very accurately performed with MRI, but competing modalities, such as single-photon emission computed tomography (SPECT) scanning and echocardiography, are more commonly used in clinical practice. Some applications, such as coronary artery imaging, are currently more accurately performed with other modalities.

(For examples of cardiac MRI scans, see the images below.)

A sagittal single-shot fast spin-echo image is use A sagittal single-shot fast spin-echo image is used as an initial localizer for coronal imaging, as shown.
Coronal single-shot fast spin-echo image obtained Coronal single-shot fast spin-echo image obtained off the sagittal plane. The image that most clearly depicts the aortic valve is selected. An oblique axial imaging plane is prescribed, as shown, from the cardiac apex to the middle of the aortic valve.

One of the main advantages of cardiac MRI is the lack of ionizing radiation, which is substantial with SPECT and computed tomography (CT) scanning. The strength of cardiac MRI, as compared to CT scanning, is its superior temporal and contrast resolution. However, the spatial resolution of CT scanning is superior.

Although competing modalities are available for every clinical application of cardiac MRI, no single modality can provide as comprehensive an evaluation as MRI. For this reason, cardiac MRI is often known as the "one-stop shop."

Technical aspects of cardiac MRI are often more daunting for the novice than are technical aspects of other modalities. The intent of this article is to serve as a primer on the technical aspects of cardiac MRI.

As the number of cardiac MRI applications is broad, the number of potential imaging techniques is correspondingly broad, and they cannot be covered thoroughly in a single article. Therefore, this article provides an in-depth review of the most commonly used cardiac MRI techniques. Several common clinical applications of cardiac MRI findings are also presented.

 

Imaging Planes

The main cardiac imaging planes are oblique to one another. As cardiac imaging planes are also at arbitrary angles with respect to the scanner, they are called “double oblique” planes. The 3 main cardiac imaging planes are the short axis, as seen in the first image below; the horizontal long axis, as seen in the second image below; and the vertical long axis, as seen in the third image below (the long axis is the line from the center of the mitral valve orifice to the left ventricular apex).

Short-axis plane: "Bright blood" steady-state grad Short-axis plane: "Bright blood" steady-state gradient echo (SSFP [steady-state free precession]) image in the short-axis plane.
Horizontal long axis: Steady-state gradient echo ( Horizontal long axis: Steady-state gradient echo (SSFP [steady-state free precession]) image in horizontal long-axis plane, also known as a 4-chamber view.
Vertical long axis: Steady-state gradient echo (SS Vertical long axis: Steady-state gradient echo (SSFP [steady-state free precession]) image in the vertical long-axis view, also known as a 2-chamber view. Mitral regurgitation is noted.

The horizontal long-axis view is also known as the 4-chamber view, and the vertical long-axis view is also known as the 2-chamber view. Note that the initial vertical long-axis view that is prescribed from an axial image is only approximate; a true vertical long-axis view should be prescribed from the horizontal long-axis view. Methods to determine the correct location and orientation of standard cardiac imaging planes have been well described.[4]

(See images below.)

A sagittal single-shot fast spin-echo image is use A sagittal single-shot fast spin-echo image is used as an initial localizer for coronal imaging, as shown.
Coronal single-shot fast spin-echo image obtained Coronal single-shot fast spin-echo image obtained off the sagittal plane. The image that most clearly depicts the aortic valve is selected. An oblique axial imaging plane is prescribed, as shown, from the cardiac apex to the middle of the aortic valve.
Four-chamber gradient-echo image (flip angle, 15°) Four-chamber gradient-echo image (flip angle, 15°) obtained off the prescribed oblique axial plane serves as a localizer for the short-axis view. As shown, the short-axis image is obtained by orienting the imaging plane perpendicular to the ventricular septum.
Coronal single-shot fast spin-echo image obtained Coronal single-shot fast spin-echo image obtained off the sagittal plane. To localize the atrioventricular valve level, an imaging plane is prescribed along a line connecting the cardiac apex and the level of the middle of the atrioventricular valve, as shown.
Four-chamber gradient-echo image (flip angle, 15°) Four-chamber gradient-echo image (flip angle, 15°) obtained off the coronal localizer provides a cross-sectional depiction of the atrioventricular valves. On the image, the right ventricle is anterior, and the left atrium and the ventricle are posterior; the mitral valve is visible in an axial plane.

Other imaging planes that may be useful include the left ventricular outflow tract view (see first image below) for ascending aortic pathology and the 3-chamber view (see second image below).[5] The 3-chamber view can be prescribed from the left ventricular outflow tract view of a short-axis view. This view displays the aortic and mitral valves immediately adjacent to one another.[6]

Left ventricular outflow view: Steady-state gradie Left ventricular outflow view: Steady-state gradient echo (SSFP [steady-state free precession]) left ventricular outflow view demonstrates a bicuspid aortic valve and an aortic jet secondary to aortic stenosis.
Three-chamber view: Steady-state gradient echo (SS Three-chamber view: Steady-state gradient echo (SSFP [steady-state free precession]) 3-chamber view. This is useful for demonstrating aortic and mitral valves on the same image. Note the close proximity of the aortic and mitral valves, which are not separated by a muscular crista, as the tricuspid and pulmonic valves are.

Unlike pulmonic and tricuspid valves, which are separated by the muscular crista supraventricularis, aortic and mitral valves are in close proximity, and often both are affected by pathologic processes.

 

Electrocardiographic Gating

Electrocardiographic gating can be performed prospectively or retrospectively.

Prospective gating

Prospective gating is most common. In prospective gating, MRI acquisition is triggered by the R wave. Within an R-R interval, a trigger delay, an acquisition window, and a trigger window may be found.[7]

In prospective gating, diastolic imaging with fast spin-echo sequences may be desirable; a trigger delay can be used to delay image acquisition after the R-wave trigger. The trigger window is the interval between completion of data acquisition and initiation of the next R wave. With a trigger window, earlier than expected heartbeats will still trigger acquisitions. The trigger window typically constitutes 10-15% of the R-R window. The acquisition window is the duration of data acquisition. A standard trigger window with no trigger delay constitutes 85-90% of the R-R window. Blood signal is optimally nulled at systole, when blood flows fastest. Because of the trigger window, prospective gating sequences exclude late diastole.

Common problems with electrocardiogram (ECG)-triggered acquisitions include poor or inaccurate R-wave detection (eg, triggering off a prominent T wave) and patient arrhythmias. R wave–detection problems can often be resolved by adjusting electrode position or by toggling the lead polarity.

Arrhythmias can result in inaccuracies in evaluation of cardiac function. Acquisition time can be increased, as some heartbeats may not trigger data acquisition. Effects of arrhythmias can be mitigated by the use of very fast sequences (eg, single-shot fast spin echo) or real-time sequences.

Retrospective gating

Retrospective gating is useful in patients with arrhythmias, because data from irregular heartbeats can be rejected.

In retrospective gating, data are acquired continuously, along with an ECG tracing. Data are retrospectively sorted by using the ECG tracing after acquisition. This is computationally intensive.

Retrospective gating is helpful for patients with arrhythmias. In retrospective gating, there is no trigger window and the full cardiac cycle is imaged. Imaging of the full cardiac cycle may result in more accurate assessment of cardiac function than is attained with prospective gating. Retrospective gating is particularly helpful when peripheral pulse gating is used. Peripheral pulse gating is an option if central gating cannot be performed. Prospectively gated peripheral pulse–triggered sequences begin after the onset of systole, as the systolic pulse must propagate to the finger before it can be detected.

 

Cardiac Imaging Sequences

Cardiac MRI imaging sequences may seem daunting to the novice. One way to approach the array of different sequences is to organize them by application.[8] Terms used in the following discussion are explained in the sections on specific sequences. It is helpful to learn the generic names of sequences rather than their trade names.[2]

It has been proposed that cardiovascular MRI pulse sequence terminology should be simplified to improve the clarity of cardiovascular MRI reports, potentially leading to greater acceptance and application of cardiovascular MRI in clincal practice.[9] The following terms have been suggested for use in clinical reports:

  • Black blood cardiac magnetic resonance (CMR)
  • Late gadolinium enhancement CMR
  • Edema CMR
  • Cine CMR
  • Strain CMR
  • Perfusion CMR
  • Flow CMR
  • Iron CMR
  • CMR coronary angiography

Cardiac function

Cardiac function is evaluated via cine gradient echo sequences often known as “bright blood” sequences (see image below). Steady-state free precession (SSFP) gradient echo sequences have largely replaced spoiled gradient echo sequences for this purpose. Trade names for these SSFP sequences include TrueFISP (True Fast Imaging With Steady-State Precession; Siemens), FIESTA (Fast Imaging Employing Steady-State Acquisition; GE), and b-FFE (Balanced Fast-Field Echo; Phillips). These sequences are typically used in conjunction with segmented k-space acquisition.

Short-axis plane: "Bright blood" steady-state grad Short-axis plane: "Bright blood" steady-state gradient echo (SSFP [steady-state free precession]) image in the short-axis plane.

Morphology

Fast spin-echo sequences, often known as “black blood” sequences, are typically used (see image below). Multiple options are available, but half-Fourier, single-shot, fast spin-echo (SS-FSE) sequences are the fastest. Trade names for these half-Fourier single-shot sequences include HASTE (Half-Fourier Acquired Single-Shot Turbo Spin Echo) and SS-FSE.

Arrhythmogenic right ventricular dysplasia: "Black Arrhythmogenic right ventricular dysplasia: "Black blood" single-shot fast spin-echo image demonstrates fatty infiltration (arrow) of the right ventricular free wall.

These sequences are typically used in conjunction with double inversion recovery prepulses. The “bright blood” SSFP sequence can also be used to assess cardiac morphology if it is altered to produce images of the entire heart (rather than a cine loop at a single location).

Perfusion

Magnetization-prepared gradient echo sequences are used to assess myocardial perfusion (see image below). The magnetization preparation prepulse can be a saturation or inversion recovery pulse and is used to improve T1-weighted contrast. Trade names for these sequences include TurboFLASH (Fast Imaging Using Low Angle Shot), Fast SPGR (Spoiled Grass [Gradient Recall Acquisition Using Steady States]), and TFE (Turbo Field Echo). Echoplanar sequences can also be used.

Myocardial perfusion: Magnetization-prepared gradi Myocardial perfusion: Magnetization-prepared gradient echo images performed to evaluate myocardial perfusion. There is normal myocardial enhancement (B) compared to the precontrast image (A).

Viability/Infarction

Contrast-enhanced MRI evaluation of myocardial viability utilizes inversion recovery gradient echo sequences, with inversion time set to null viable myocardium. Spoiled gradient echo or SSFP sequences can be used in conjunction with the inversion recovery prepulse. These sequences typically utilize segmented k-space acquisition.

Flow/Velocity

Flow quantification utilizes cine phase contrast sequences (see images below).

Aortic velocity vs time plot from the phase contra Aortic velocity vs time plot from the phase contrast acquisition in the image below.
Axial phase contrast acquisition. The bright signa Axial phase contrast acquisition. The bright signal is secondary to flow in the ascending aorta.

Angiography

Many different sequences have been used to image the coronary arteries. These sequences are typically used in conjunction with segmented k-space acquisition. Two-dimensional (2D), segmented, gradient echo sequences can be used to evaluate coronary artery anomalies. Three-dimensional (3D) techniques are used to evaluate the arteries for stenosis. Images can be acquired during breath-holding or during free breathing and can be obtained with or without intravenous contrast. A 3D, segmented SSFP sequence without intravenous contrast is well suited for evaluation of the coronary arteries.[10] If intravenous contrast is employed, intravascular contrast agents are most useful.

Standard 3D, spoiled gradient echo sequences with intravenous contrast are used to evaluate the aorta and the great vessels.

 

Morphology: Black Blood Sequences

The following discussion covers the most commonly used sequences in the clinical practice of cardiac MRI: cine SSFP, fast spin echo with double inversion recovery, and inversion recovery gradient echo. A conceptual approach is used to explain why specific techniques and sequences are used.

Black blood MRI scans (see image below) are produced with sequences designed to null the signal of flowing blood. These images allow anatomic assessment of the heart and vascular structures without interference from a bright blood signal. Although black blood sequences are standard in most imaging protocols, they are particularly important for assessing cardiac masses, the myocardium (eg, in suspected arrhythmogenic right ventricular dysplasia), and the pericardium.

Arrhythmogenic right ventricular dysplasia: "Black Arrhythmogenic right ventricular dysplasia: "Black blood" single-shot fast spin-echo image demonstrates fatty infiltration (arrow) of the right ventricular free wall.

In clinical practice, 3 general options are available for black blood imaging.[11]  The first 2 options are used most often.

  • Half-Fourier, single-shot fast spin echo with double inversion recovery

  • Breath-hold, single-slice fast spin echo with double inversion recovery

  • Multislice fast spin echo

How long does black blood MRI take? In ECG-gated spin-echo cardiac imaging, TR (repetition time) depends upon heart rate or R-R interval. Thus, acquisition time can be calculated by substituting the R-R interval for the TR in this standard equation:

  • Acquisition time = R-R interval × number of phase-encoding steps × number of acquisitions/echo train length

Note that if the heart rate is 70 beats/min, the R-R interval is 857 msec, which may not be adequate for T2-weighted imaging. In this case, triggering can be performed after every other R wave, and "2 × R-R interval" should be used in place of "R-R interval" in the above equation.

Should I use T1-weighted, proton density, or T2-weighted images?

In many cases, the purpose of black blood imaging is to assess anatomy, and weighting is not important. In such cases, TR should be as short as possible to minimize imaging time; thus, black blood MRI scans are often T1-weighted. For some applications, such as cardiac mass evaluation, specific T2-weighted sequences may be performed.

Why is the blood black?

Protons must experience the 90° excitation pulse and the 180° refocusing pulse to generate a spin echo. If protons in flowing blood are not present in the slice long enough to experience both pulses, no spin echo is generated.

Thus, a way to minimize the signal from flowing blood is to decrease the chance that flowing blood will experience 90° and 180° pulses. This can be done by minimizing the time the blood is in the slice—for example, by decreasing the volume of the slice (thinner slices), creating the shortest path (slice positioning orthogonal to flowing blood), or increasing the speed of flowing blood (imaging during systole). Another way to minimize the signal from flowing blood involves increasing the time interval between the 90° and 180° pulses (TE, or echo time).

In standard spin-echo imaging, acquisition during systole results in increased nulling of the blood signal. However, as will be discussed, in fast spin-echo imaging (see below), diastolic imaging is usually optimal.

Fast spin-echo imaging

Standard spin-echo black blood imaging has little utility in clinical practice because acquisition times exceed patient breath-holding times. Although resulting respiratory artifacts can be remedied to some extent with signal averaging (which further increases acquisition time), acquisition during free breathing is better performed with multislice fast spin-echo imaging. The fastest fast spin-echo sequences can be performed during a breath-hold.

A basic disadvantage of fast spin-echo imaging relative to spin-echo imaging is the image blurring that results from acquiring data at different effective echo times during the echo train. In cardiac imaging, this image blurring is exacerbated by increased motion at systole. Thus, to minimize artifact, fast spin-echo cardiac MRI is best performed in diastole.

Fast spin-echo cardiac MRI sequences are typically performed, and double inversion recovery (see below) pulses are added to achieve optimal nulling of blood signal.

Double inversion recovery

Double inversion recovery sequences are designed specifically to null the signal from flowing blood. There are 2 prepulses. A nonselective 180° radiofrequency (RF) pulse inverts all protons. This is followed by a slice-selective 180° pulse that reverts all protons in the imaging slice back to their original alignment. This results in no effect on stationary protons in the imaging slice. However, flowing blood in the imaging slice will have experienced only the nonselective pulse (the blood that experienced both pulses will no longer be in the slice at the time of imaging). Double inversion recovery sequences begin imaging when the magnetization vectors of flowing blood cross the null point—the inversion time.

Typical inversion times for double inversion recovery sequences are between 400 and 600 msec and depend on heart rate. Note that inversion time accounts for a substantial portion of a typical R-R window, which limits the time available to acquire the echo train. Also note that imaging begun 400 to 600 msec after initiation of the R wave will conveniently be performed in diastole.

What sequence should I use?

The fastest sequences are half-Fourier, single-shot fast spin echo with double inversion recovery, in which data needed to generate an image can be acquired during a single heartbeat. However, although these images have the least cardiac and respiratory motion artifact, the half-Fourier single-shot acquisition decreases spatial resolution and the signal-to-noise ratio (SNR). For applications in which optimal resolution and signal are useful (eg, evaluation of the right ventricular wall in suspected arrhythmogenic right ventricular dysplasia), breath-hold, single-shot fast spin echo with double inversion recovery (1 slice per breath-hold) may be more useful.

Another option is to use multislice fast spin-echo imaging during free breathing. This technique is similar to basic spin-echo imaging, but a short-echo train is added to decrease imaging time. As in spin-echo sequences, multiple signal averaging is used to decrease respiratory motion artifact. As blurring is minimal with a short-echo train, systolic imaging is possible and blood nulling is similar to use of spin-echo sequences. Inversion recovery pulses may not be necessary with this technique.

It is possible to use bright blood sequences to evaluate cardiac morphology.

 

Function: Bright Blood Sequences

Steady-state gradient echo imaging has largely replaced spoiled gradient echo imaging for bright blood cine cardiac MRI (see image below).

Short-axis plane: "Bright blood" steady-state grad Short-axis plane: "Bright blood" steady-state gradient echo (SSFP [steady-state free precession]) image in the short-axis plane.

Steady-state versus spoiled gradient echo imaging

In gradient echo (GRE) imaging, the TR is often shorter than the T2 of most tissues, and transverse magnetization will not have fully decayed before the next RF pulse. Thus, residual transverse magnetization will add T2 contrast (in addition to T1 contrast) to the image. This additional T2 contrast is undesirable for many applications, as T1 and T2 contrast may be competitive. For example, a liver lesion that is hypointense on T1 and hyperintense on T2 may be isointense with both T1 and T2 weighting. To achieve T1 weighting with a short TR GRE sequence, spoiling the residual transverse magnetization is necessary. This spoiling can be accomplished with an RF pulse or gradient. A majority of fast GRE sequences used in noncardiac clinical MRI are spoiled.

In steady-state GRE sequences, spoiling is not performed and residual transverse magnetization is retained. Retained residual transverse magnetization increases the SNR of steady-state sequences relative to spoiled sequences. Image contrast will depend on the T2-to-T1 ratio. As has been stated, this is undesirable for many applications. In steady-state sequences, only fluid and fat will have a high signal (fluid and fat have comparable T1 and T2 times, but in most other tissues, T2 time is much shorter than T1 time). However, in bright blood cardiac MRI, hyperintense blood relative to other tissues is exactly what is needed; thus, steady-state GRE sequences are optimal for cine cardiac imaging (cMRI).

The sequences used in cardiac imaging are balanced SSFP sequences. Trade names for these sequences include TrueFISP, FIESTA, and balanced FFE. These sequences are very fast and have a high SNR, but T2-to-T1 image contrast limits the role of these sequences in noncardiac applications.

SSFP cine MRI has largely replaced spoiled GRE cine MRI for evaluation of cardiac function. SSFP sequences do not depend on flow; they have a higher SNR and are faster. Spoiled GRE sequences are T1-weighted and depend on through-plane flow enhancement (similar to time-of-flight MR angiography) to generate contrast. The blood may become saturated if flow is slow or if TR is short. Spoiled GRE cine MRI does not allow for use of very low TRs because there is not enough time for saturated blood to be replaced by unsaturated blood between excitation pulses.

With SSFP sequences, blood signal is dependent on intrinsic contrast rather than on inflow effects, and TR can be as short as possible. SSFP cine MRI can be almost 3 times as fast as spoiled GRE cine MRI. In addition, the SSFP sequence has a higher SNR because of residual transverse magnetization. This is particularly true at low TRs. With spoiled GRE sequences, SNR is reduced with decreasing TR. With SSFP sequences, SNR is high even at low TRs because residual transverse magnetization is increased with shorter TRs.

Requirements for SSFP imaging

High-quality SSFP imaging depends on a low TR, a high flip angle, and a uniform magnetic field.[10]

In SSFP imaging, residual transverse magnetization must be preserved. Field inhomogeneity and unbalanced gradients can disrupt steady-state transverse magnetization. Sequences are implemented with balanced gradients to minimize gradient-induced dephasing. SSFP sequences are very sensitive to field inhomogeneities. In regions of high local magnetic field variation, SSFP images often suffer from characteristic bands of signal loss (off-resonance banding artifact), which can disrupt the steady state.

As TR is increased, any off-resonance banding artifact will become more pronounced because of increased off-resonance precession per TR. Thus, the lowest TR possible is desirable for SSFP imaging. Typical TRs are less than 4 msec, with TEs less than 2 msec. Banding artifact is a particular limitation for 3D MRI, as the banding artifact becomes more pronounced as the main magnetic field strength (and any associated inhomogeneity) is increased.

In spoiled GRE sequences, optimal SNR is dependent on matching the flip angle to the TR (the lower the TR, the lower the flip angle). In SSFP sequences, the SNR does not change substantially with different flip angles, but T2/T1 weighting is increased with an increasing flip angle. SSFP sequences should use the largest flip angle achievable because this will maximize the contrast-to-noise ratio. As RF pulses are continuously applied to maintain steady state, specific absorption rate limits are often a factor in SSFP sequences and limit the use of very high flip angles. Flip angles in SSFP sequences are typically 40º to 70°.

Limitations of SSFP imaging

SSFP sequences are prone to off-resonance banding artifacts. As these artifacts are caused by local field inhomogeneities, a very uniform magnetic field is required to avoid artifacts.[12]

Because SSFP sequences are typically performed with very low TRs and TEs, low TE time may result in a chemical shift artifact of the second kind (India ink artifact).

SSFP sequences (see first image below) may be less sensitive to turbulent flow (eg, in regurgitant valves) compared to spoiled GRE sequences (see second image below) because SSFP sequences do not depend on time-of-flight effects.

Aortic regurgitation (steady-state gradient echo): Aortic regurgitation (steady-state gradient echo): Steady-state gradient echo (SSFP [steady-state free precession]) images demonstrate turbulent flow secondary to aortic regurgitation. Compared with spoiled gradient echo images, the dephasing secondary to turbulent flow is less well visualized. However, overall contrast is greater (eg, delineation of endocardial borders).
Aortic regurgitation (spoiled gradient echo): Spoi Aortic regurgitation (spoiled gradient echo): Spoiled gradient echo images demonstrate turbulent flow secondary to aortic regurgitation. Compared with steady-state gradient echo images, the dephasing secondary to turbulent flow is better visualized, but overall image contrast is less (eg, delineation of endocardial borders).
 

Function: Temporal Resolution, Spatial Resolution, and Imaging Time

Temporal resolution and frames

Multiple images at the same slice position, corresponding to different time points in the cardiac cycle, are obtained during cine GRE imaging. Each image is called a frame. Typically, 12 to 18 frames are obtained during a cardiac cycle. Temporal resolution is the duration of the cardiac cycle that each frame represents. High temporal resolution is necessary to accurately assess cardiac motion, particularly during systole.[12]

The ideal temporal resolution should be 50 to 60 msec or less. With faster heart rates, greater temporal resolution is needed.[12]

Temporal resolution and number of frames are directly related, but in general, temporal resolution is more important than the number of frames obtained. For example, with a very fast heartbeat, functional evaluation may still be adequate with good temporal resolution, although fewer frames may be obtained than are typically obtained with slower heartbeats.[12]

Segmented k-space cine GRE

In cine MRI, echoes are partitioned into k-spaces, with each k-space corresponding to a frame. With 12 frames, echoes would be partitioned into 12 k-spaces. The quantity of data (number of phase-encoding steps) needed to fill each of the k-space corresponds to spatial resolution. In conventional cine MRI, each of the 12 k-spaces is filled with only 1 phase-encoding step of necessary data during a single heartbeat. Total acquisition time is therefore the number of heartbeats necessary to fill a k-space.

A standard study with 128 phase-encoding steps will take 128 heartbeats to complete; this does not allow for breath-hold imaging. With segmented k-space cine MRI, multiple phase-encoding steps of data (per frame) are acquired after a single heartbeat.

The number of lines of k-space per frame acquired per heartbeat is referred to as views per segment or lines per segment. For a study with 128 phase-encoding steps, 8 views per segment would reduce imaging time from 128 heartbeats to 16 heartbeats. This would allow for breath-hold cardiac cine imaging.

Relationship between temporal resolution, spatial resolution, and imaging time

An understanding of the relationship between temporal resolution, spatial resolution, and imaging time is important in cine cardiac MRI.[12]

Temporal resolution is directly related to views per segment:

  •  Temporal resolution = TR × views per segment

In this case, TR is used in the standard sense to refer to time between consecutive RF pulses. Lee refers to this as “true TR” because TR is also used to refer to temporal resolution in cine MRI.[12]

There is a direct trade-off between imaging time (views per segment) and temporal resolution. Reducing imaging time by increasing views per segment decreases temporal resolution. For example, when the number of views per segment is doubled, overall imaging time is decreased by half because twice the quantity of data is acquired during every heartbeat. However, acquiring twice the data per heartbeat takes twice as long per frame; this halves the number of attainable frames per cardiac cycle and worsens temporal resolution by a factor of 2.

Another way to shorten imaging time is to reduce resolution by decreasing the number of phase-encoding steps. In-plane spatial resolution of 2 to 2.5 mm is adequate for most cardiac function studies, although higher spatial resolution can be helpful for evaluating structures such as cardiac valves. In patients who are poor breath-holders, temporal resolution, spatial resolution, or both, must be compromised when scan time needs to be decreased.

Heart rate can be helpful in determining the number of views per segment. With a slow heart rate, more views per segment can be used. Because the R-R interval is longer, more views per segment can be added while an adequate number of frames are maintained, but temporal resolution is still decreased. This decreases the number of heartbeats required to complete the study, which is especially helpful if heart rate is slow.

Other techniques

Temporal resolution can be increased with minimal effect on acquisition time through view sharing or echo sharing. In echo sharing, echoes are recycled over multiple images; this can improve perceived temporal resolution.

Discussion of parallel imaging techniques is beyond the scope of this article, but parallel imaging is especially useful in conjunction with SSFP cine cardiac imaging. Parallel imaging techniques can reduce imaging time substantially by decreasing the number of phase-encoding steps necessary to reconstruct an image severalfold. The drawback is decreased SNR, but SSFP cine MRI has inherently high SNR and is therefore more tolerant of decreased SNR resulting from parallel imaging.

 

Function: Ventricular Analysis

Many methods of ventricular volume calculation are used. Rehr and Putnam found excellent correlation between findings at volumetric analysis with MRI and findings with ventricular casts (0.99 correlation; 4.9 mL standard error).[13] Accuracy is increased by inclusion of long-axis measurements. Three-dimensional volumetric calculations are well correlated with ventriculographic findings and have low interstudy variability (< 5%), as compared to ventriculographic and echocardiographic results.

The first step in calculation of ventricular volume consists of selection of representative end-diastolic (ED) and end-systolic (ES) cardiac-phase images. According to Semelka et al, phase images that depict the largest and the smallest ventricular volumes or phase images obtained immediately before mitral valve closure (ie, ED) and opening (ie, ES) are chosen. Next, the right ventricle (RV) and the left ventricle (LV) are traced along the endocardial margin on each section obtained in selected ED and ES phases from the cardiac apex to the section just prior to the one that depicts the mitral and tricuspid valves.[14]

The operator can make some choices when tracing ventricular margins.[12] One choice involves whether to include trabeculations and papillary muscles. It is most important that the technique be consistent between studies. In patients with hypertension, hypertrophic cardiomyopathy, or storage disease, papillary muscles may hypertrophy; therefore, inclusion of papillary muscles may yield the most accurate quantification of myocardial mass. For evaluation of function, trabeculations and papillary muscles can be included or excluded, as long as the technique is consistent.

Because the long-axis dimension of the left ventricle is shorter during systole, the basal short-axis slice may include the left atrium during systole and the left ventricle during diastole. If short-axis images are used, the operator may need to decide which slices should be included at the left ventricular base. Usually, only slices on which a complete circumferential left ventricular rim is visualized should be included.

ED and ES volumes for each section are totaled to yield RV and LV end-diastolic volume (EDV) and end-systolic volume (ESV). When short-axis images are used, volumes can be calculated using Simpson's rule: sum of cross-sectional areas of each slice × distance between slices.

  • Stroke volume (SV) = EDV – ESV.

Ejection fraction (EF) equals SV divided by EDV times 100, or EF = (SV/EDV) × 100, to yield a value reported as a percentage.

  • Cardiac output = SV × heart rate.

For myocardial mass assessment, RV and LV epicardial borders are traced in ED. The interventricular septum is assigned to the LV and is excluded from the RV tracing of myocardial mass (see images below). The volumes of all sections are added and the corresponding EDV is subtracted to determine myocardial volume. This result is then multiplied by specific gravity of the myocardium (ie, 1.05 g/mL) to calculate the mass. This measurement is useful for assessing hypertrophy and in following up on ventricular response to therapy.

Short-axis gradient echo image (flip angle, 15º) i Short-axis gradient echo image (flip angle, 15º) illustrates volumetric analysis, with tracing of the endocardial border and exclusion of the papillary muscle superiorly. The epicardial margin also is traced for calculation of the myocardial mass, in which the septal mass is assigned as part of the left ventricle.
Short-axis gradient echo image (flip angle, 15°) o Short-axis gradient echo image (flip angle, 15°) of the ventricles illustrates volumetric calculation of the right ventricle, with a tracing of the right ventricular endocardial border. The epicardium also is traced for determination of the myocardial mass. Note that the septum is excluded when the epicardium is traced.

Atrioventricular and ventriculoarterial valves can be assessed via cine GE sequences.[15] Valvular stenosis or regurgitation produces turbulent jets of signal void in appropriate directions. Regarding AV valves, regurgitation is graded according to echocardiographic criteria and is related to the distance by which the jet extends into the atrium. Grades of valvular stenosis are calculated more reproducibly. Valve orifice areas can be measured and graded according to current standards (see image below).

Short-axis gradient echo end-diastolic image (flip Short-axis gradient echo end-diastolic image (flip angle, 20°) of the outflow tract shows the open aortic valves. The valve area can be measured in this manner.

It is important to note that the MRI sequence used may affect the calculated ventricular volume and mass,[16] which is dependent on accurate delineation of endocardial and epicardial borders. Delineation of the endocardial border is dependent on contrast between the myocardium and the ventricular blood pool. SSFP images (see image below) show greater contrast between the myocardium and the blood pool than do spoiled GRE images, primarily because the blood pool signal is not dependent upon flow.

Aortic regurgitation (steady-state gradient echo): Aortic regurgitation (steady-state gradient echo): Steady-state gradient echo (SSFP [steady-state free precession]) images demonstrate turbulent flow secondary to aortic regurgitation. Compared with spoiled gradient echo images, the dephasing secondary to turbulent flow is less well visualized. However, overall contrast is greater (eg, delineation of endocardial borders).

On spoiled GRE images (see image below), poor delineation of the endocardial border may result in an artifactual increase in apparent myocardial thickness. When SSFP sequences are used, ventricular volumes are higher and myocardial mass is lower than values derived from spoiled GRE sequences.

Aortic regurgitation (spoiled gradient echo): Spoi Aortic regurgitation (spoiled gradient echo): Spoiled gradient echo images demonstrate turbulent flow secondary to aortic regurgitation. Compared with steady-state gradient echo images, the dephasing secondary to turbulent flow is better visualized, but overall image contrast is less (eg, delineation of endocardial borders).

Normal LV volumes and masses (Table 1) differ between sexes and among certain ethnic groups.[17] Men have significantly higher LV volumes and larger masses than women. Asian Americans, regardless of sex, have lower LV volumes and smaller mass than persons of other ethnic groups. African American men have the largest LV volumes and mass. LV volumes do not differ between African American women and white or Hispanic American women. These findings are significant after normalization for body surface area. LV mass is independent of age if indexed for body surface area.

Table 1. Left Ventricular Parameters [17] (Open Table in a new window)

Parameter

Men

Women

LV EDV (mL)

142 ± 34

109 ± 22

LV ESV (mL)

47 ± 19

31 ± 9

LV EF (%)

67 ± 7

72 ± 6

LV SV (mL)

95 ± 21

78 ± 17

Cardiac output (mL/min)

5.6 ± 1.2

4.9 ± 1.1

LV mass (g)

164 ± 36

114 ± 24

EDV: end-diastolic volume; EF: ejection fraction; ESV: end-systolic volume; SV: stroke volume

Valvular parameters

Typical areas for aortic and mitral valves are 2.5 to 3.5 cm2 and 4 to 6 cm2, respectively; areas less than 0.8 cm2 and less than 1 cm2, respectively, indicate severe stenosis. Values for aortic and mitral valve areas apply to both males and females.

 

Infarcts and Viability: Contrast-Enhanced Inversion Recovery GRE

Delayed enhancement of infarcts distinguishes infarct from viable myocardium. The difference in intensity between infarcted myocardium and viable myocardium may be difficult to detect on T1-weighted images. GRE sequences can be performed with an inversion recovery prepulse to optimize visualization of enhancing infarcted myocardium. Inversion time is selected to null signal from viable myocardium.

Typical correct inversion times are between 200 and 300 msec.[11] The correct inversion time can be determined empirically via an inversion-time mapping sequence (eg, a segmented cine GRE [typically, SSFP] with images generated at multiple inversion times). This process has been called “inversion-time surfing.” The SSFP sequence is optimal for this technique because longitudinal magnetization is minimally disturbed during readout.

A complicating factor is that nulling inversion time will slightly increase during the course of the exam.[18] As gadolinium washes out of viable myocardium, inversion time is increased.

When inversion time is optimal, the normal myocardium should be very dark (see image below), the infarct should be the most intense structure on the image, and the left ventricular blood pool should be of intermediate intensity.[19]

Myocardial infarct: Delayed contrast-enhanced inve Myocardial infarct: Delayed contrast-enhanced inversion recovery gradient echo image demonstrates delayed enhancement of the apex and anterolateral wall (arrow) consistent with scar. Note the hypointensity of the normal myocardium, suggesting that an optimal inversion time was selected.

If inversion time is too short, the blood pool will appear dark. If inversion time is close to the optimal value but is still short, the myocardium may have a speckled appearance and endocardial and epicardial borders may appear as hypointense lines around intermediate-intensity myocardium. If inversion time is too long, the area of delayed enhancement will be only slightly more intense than the myocardium (see image below).

Transmural and nontransmural infarcts: Delayed con Transmural and nontransmural infarcts: Delayed contrast-enhanced inversion recovery gradient echo image demonstrates delayed enhancement in the anterior (arrowhead) and inferior (arrow) walls. The full thickness of the inferior wall enhances consistent with transmural infarction. There is only partial-thickness enhancement of the anterior wall consistent with nontransmural infarction. The percentage of wall thickness demonstrating delayed enhancement negatively correlates with recovery of function after revascularization. Note that the degree of hypointensity in the normal myocardium is less than that seen in the image above, suggesting that the inversion time chosen was not optimal in this case.

One way to minimize the importance of choosing inversion time correctly is to use phase-sensitive reconstruction.[11] Unlike images of typical magnitude, which do not distinguish between signal from protons above and those below the x-y plane, signal intensity in phase-sensitive images varies with longitudinal magnetization across the full spectrum. It is much less important to accurately choose T1 because infarcted myocardium should always be higher in signal intensity than viable myocardium at all inversion times.

The disadvantage of phase-sensitive reconstruction is that background noise that has a random phase is pixilated. If inversion time is correctly chosen, magnitude images may be preferable.

The contrast-enhanced inversion recovery GRE sequence is performed with k-space segmentation. A spoiled GRE or SSFP sequence can be used. Typically, a breath-hold technique is used with 1 section per breath-hold and 10 to 12 sections total. If the patient can sustain a longer breath-hold, a single 3D acquisition can cover the entire heart. The sequence can also be performed with free breathing and respiratory gating.

 

Applications

Cardiac MRI has become an indispensable imaging modality in the investigation of patients with suspected heart disease. It has emerged as the gold standard test for cardiac function, volumes, and mass and allows noninvasive tissue characterization and assessment of myocardial perfusion. Quantitative MRI already has a key role in the development and incorporation of machine learning in clinical imaging, potentially offering major improvements in both workflow efficiency and diagnostic accuracy. As clinical applications of a wide range of quantitative cardiac MRI techniques are being explored and validated, we are expanding our capabilities for earlier detection, monitoring, and risk stratification of disease, potentially guiding personalized management decisions in various cardiac disease models.[20]

Following is a brief overview of several common clinical applications of cardiac MRI. With an understanding of the technical aspects specific to cardiac MRI, the imager should be able to optimize image quality to aid in diagnosis and management of these and other cardiovascular diseases.

Ischemic heart disease

Ventricular function can be quantitatively assessed by reviewing dynamic images. Chronic transmural infarcts exhibit lack of wall thickening in systole and reduced myocardial thickness (< 6 mm).[19] Spin-echo images may show an area of decreased signal intensity that corresponds to postinfarct scar formation.

Myocardial tagging is used to track segmental motion and can help in directly distinguishing impaired myocardium from myocardium that may move abnormally because of its proximity to a disease area (ie, tethering effect). Presaturation pulses are used to create a cross-hatched pattern over the myocardium. As systole proceeds, the pattern is distorted in a direction that corresponds with myocardial movement (see images below). Thus, myocardial contraction can be assessed reproducibly and quantitatively.

Short-axis gradient echo end-diastolic image (flip Short-axis gradient echo end-diastolic image (flip angle, 10°) shows the crosshatched tagging lines, which are regularly spaced over the left ventricular wall.
Short-axis gradient echo image (flip angle, 10°) o Short-axis gradient echo image (flip angle, 10°) obtained in end systole shows distortion of the crosshatch pattern. The distortion is symmetric, with bowing of all transverse lines toward the center, which indicates lack of wall motion abnormality. The section is toward the apex, and the radial lines show counterclockwise displacement.

Areas lacking pattern distortion indicate nonfunctioning myocardium. Experience with tagging has elucidated the characteristics of cardiac dynamics.[21] Ventricular contraction contains a twisting component in addition to short- and long-axis movement. A wringing effect occurs, whereby the base moves clockwise and the apex moves counterclockwise. This torsion is reversed during isovolumic relaxation and before the mitral valve is opened. This mechanism may generate ventricular suction, promoting early diastolic filling. In addition, long-axis shortening is pronounced in lateral and posterior walls, as opposed to that in anterior and septal regions.

The ultimate focus of MRI in this application is the development of techniques for identifying acute infarction and for differentiating viable myocardium from nonviable myocardium. The literature is filled with reports of MRI studies of acute myocardial infarction (MI).

Many articles discuss the presence of increased signal intensity on T2-weighted images of regions with acute infarction; Filipchuk et al revealed that although sensitivity was adequate (88%), specificity was only 17% as compared to that of controls. Myocardial thinning was the most specific finding in MI (88%), and sensitivity was only 67%.[22] Subendocardial signal intensity changes can be difficult to distinguish from flow-related enhancement.

A more reliable indicator of acute MI is delayed contrast enhancement after IV administration of gadolinium-based contrast material. Studies in patients after MI found good correlation between increased enhancement and infarcted tissue compared with patients with healthy myocardium.

After the contrast has washed out of other regions (5-10 minutes), it is retained in altered cells after MI. In normal myocardium, gadolinium is excluded from the myocyte intracellular space. Sarcolemmal membrane integrity is lost in cell death, allowing gadolinium to extravasate into the myocyte, resulting in hyperenhancement.

In the chronic setting, scar tissue has increased extracellular collagen and a larger interstitial space than normal myocardium. A larger interstitial space would account for delayed hyperenhancement seen in scar. Thus, T1-sensitive inversion recovery imaging set to null (blacken) normal myocardium produces a "scar map." Size and transmural depth are slightly greater shortly after MI, but from 1 week onward, they are stable indicators of scar extent.

Delayed hyperenhancement is associated with myocyte necrosis in the setting of acute and chronic infarction. "At risk" myocardium and severe reversible ischemic injury (even in the setting of stunning) do not exhibit hyperenhancement. A greater transmural extent of infarction (eg, hyperenhancement involving >50% of wall thickness) can predict regions that are less likely to improve in function after revascularization or beta-blocker therapy (see image below). The extent of dysfunctional but nonhyperenhanced myocardium can predict improvement in LV EF after therapy.[23]

Myocardial infarct: Delayed contrast-enhanced inve Myocardial infarct: Delayed contrast-enhanced inversion recovery gradient echo image demonstrates delayed enhancement of the apex and anterolateral wall (arrow) consistent with scar. Note the hypointensity of the normal myocardium, suggesting that an optimal inversion time was selected.
Transmural and nontransmural infarcts: Delayed con Transmural and nontransmural infarcts: Delayed contrast-enhanced inversion recovery gradient echo image demonstrates delayed enhancement in the anterior (arrowhead) and inferior (arrow) walls. The full thickness of the inferior wall enhances consistent with transmural infarction. There is only partial-thickness enhancement of the anterior wall consistent with nontransmural infarction. The percentage of wall thickness demonstrating delayed enhancement negatively correlates with recovery of function after revascularization. Note that the degree of hypointensity in the normal myocardium is less than that seen in the image above, suggesting that the inversion time chosen was not optimal in this case.

In clinical practice, imaging is typically performed 10 to 15 minutes after contrast injection to detect delayed hyperenhancement. An imaging window of 10 to 30 minutes is probably acceptable. In acute infarct, hyperenhancement could potentially overestimate infarct size if a long delay is not used.[10]

Delayed hyperenhancement can be seen in diseases other than acute and chronic infarcts.[23] These include sarcoidosis, dilated and hypertrophic cardiomyopathy, myocarditis, amyloidosis, and arrhythmogenic right ventricular dysplasia. In many cases, the pattern of enhancement in these other disorders is different from subendocardial or transmural hyperenhancement seen in infarct. For example, enhancement may have a midwall, epicardial, or global endocardial distribution.

Enhancement patterns have been explored with combined first-pass and delayed imaging to assess viability and to predict functional recovery.

In the emergency department, cardiac MRI may aid in identification of patients with non–ST-segment elevation MI or unstable angina or acute coronary syndrome (ACS) with unobstructed coronary artery disease, if the patient's clinical history is known to be atypical. Also, cardiac MRI is excellent for risk stratification of patients for adverse left ventricular remodeling or major adverse cardiac events. Cardiac MRI should be performed early in the course of the disease (< 2 wk after onset of symptoms). SSFP T2-weighted MRI with late gadolinium enhancement is the mainstay of the cardiac MRI protocol. Further sequences can be used to analyze different pathophysiologic subjacent mechanisms of disease, such as microvascular obstruction or intramyocardial hemorrhage. Finally, cardiac MRI may provide several prognostic biomarkers that help in follow-up of these patients.[24]

Gadolinium-based contrast agents have been linked to the development of nephrogenic systemic fibrosis (NSF) or nephrogenic fibrosing dermopathy (NFD). NSF/NFD, a debilitating and sometimes fatal disease, has occurred in patients with moderate to end-stage renal disease after they were given a gadolinium-based contrast agent to enhance MRI or magnetic resonance angiography (MRA) scans. Characteristics include red or dark patches on the skin; burning, itching, swelling, hardening, and tightening of the skin; yellow spots on the whites of the eyes; joint stiffness with trouble moving or straightening the arms, hands, legs, or feet; pain deep in the hip bones or ribs; and muscle weakness.

Cardiac masses

Echocardiography is usually the initial step in evaluation of cardiac masses. However, MRI is better for assessing the relationship of the mass to cardiac structures, and MRI provides more reliable indications of the histologic diagnosis. A particular advantage of MRI is that it can be used to distinguish thrombus from tumor. On spin-echo images, thrombus and tumors can have intermediate signal intensity. Unlike a tumor, a thrombus has low signal intensity on GE images because of the presence of deoxyhemoglobin.[25, 26, 27, 28]

A diagnostic dilemma may arise in the differentiation of thrombus from myxoma (see images below), which can appear hypointense because of its calcification and high hemoglobin content secondary to hemorrhage.

Lipomatous hypertrophy of the interatrial septum: Lipomatous hypertrophy of the interatrial septum: Steady-state gradient echo image (SSFP [steady-state free precession]) demonstrates lipomatous hypertrophy of the interatrial septum. Note the prominent visualization of the fossa ovalis (arrowhead), which is a common finding, as lipomatous hypertrophy typically spares the fossa ovalis. This could be misinterpreted as an atrial septal defect if the lipomatous hypertrophy is not noted. In addition, an atrial thrombus (arrow) is noted (see also image below). A myxoma would be in the differential diagnosis, but a myxoma classically arises from the interatrial septum in the region of the fossa ovalis.
Right atrial mass: Contrast-enhanced spoiled gradi Right atrial mass: Contrast-enhanced spoiled gradient echo image demonstrates lack of enhancement of a right atrial mass (arrow) consistent with thrombus.

Myxoma usually originates from the interatrial septum; it can be pedunculated or sessile, and its contour is mostly smooth. Thrombus occupies the atrial appendage, is broad based, and has an irregular contour. Myxoma can prolapse through the mitral valve at cine GE imaging, whereas thrombus usually is associated with mitral valve disease. However, both can originate from the posterior atrial wall (in a minority of patients). Another pitfall in thrombus detection occurs when a hyperacute clot has increased signal intensity on GE images, which can be confused with the appearance of a tumor.

Any clinical question regarding differentiation of tumor and thrombus can be answered by administration of contrast material, which causes enhancement of the tumor but not of the thrombus. The presence of a clot in the right atrium is suggestive of a tumor thrombus and should prompt further exploration. The most likely source is renal, hepatic, or adrenal.

Other primary cardiac tumors exist. Their signal intensity characteristics are summarized in Table 2.

Table 2. Cardiac Tumor Characteristics (Open Table in a new window)

Mass

T1-Weighted MRI

T2-Weighted MRI

Enhancement

Distribution and Features

Thrombus

Isointensity

Hypointensity

None

Atrial appendage, signal void on GE images

Myxoma

Hypointensity or isointensity

Hyperintensity (often heterogeneous)

Mild or moderate

Atrial septum, in women aged 40-60 y, hypointense on GE images

Fibroma

Isointensity or hyperintensity

Isointensity or hypointensity

Rim

Most patients < 10 y, anterior wall of the left ventricle and/or septum, cystic or calcified, associated with Gorlin syndrome

Rhabdomyoma

Isointensity

Isointensity

Mild or none

Most patients < 1 y, multiple tuberous sclerosis present in 50%

Hemangioma

Isointensity

Isointensity

High

Mostly intramural but can be exophytic and polypoid

Pheochromocytoma

Hypointensity

Extreme hyperintensity

High

Usually juxtacardiac, chromocytoma, or pericardial mass

Lymphoma

Hypointensity

Hyperintensity

Heterogeneous

No data

Malignant fibrous histiocytoma (MFH)

Heterogeneous intensity

Hyperintensity

Moderate

Posterior part of the left atrium, multiple in two-thirds of patients

Angiosarcoma

Heterogeneous intensity

Heterogeneous intensity

Heterogeneous or high

Frondular, in men aged 20-50 y, extension into the great vessels

Adapted from Martin DR, Merchant N, MacDonald C. MR imaging of cardiac masses: a review of current application and approach. Appl Radiol. 2000;Mar:10-20.

Another advantage of MRI is its ability to depict the characteristics of a pseudomass. Many normal structures can simulate a tumor or a thrombus on echocardiography, and in many cases, MRI can facilitate diagnosis. A prominent crista terminalis in the right atrium (see first image below) or a moderator band in the right ventricle (see second image below) can be misinterpreted as a lesion. These can be shown to be normal anatomic structures.

Four-chamber double inversion recovery image: CT i Four-chamber double inversion recovery image: CT indicates the crista terminalis, which, when prominent, can be misdiagnosed as a cardiac tumor; RA, right atrium.
Four-chamber, double inversion recovery image: MB Four-chamber, double inversion recovery image: MB indicates the moderator band, which can be misdiagnosed as a tumor if it is thickened; LV, left ventricle; RV, right ventricle.

Large trabeculations and papillary muscles can be problematic, and asymmetric ventricular hypertrophy can simulate tumor as well. Cine GE imaging findings can confirm normal contraction and function of these structures, which are not observed with a tumor.

Furthermore, obese, female, and older patients have a predilection for lipomatous hypertrophy of the myocardium. These can occur in any location, but a classic location is the interatrial septum (see image below), typically sparing the fossa ovalis.

Lipomatous hypertrophy of the interatrial septum: Lipomatous hypertrophy of the interatrial septum: Steady-state gradient echo image (SSFP [steady-state free precession]) demonstrates lipomatous hypertrophy of the interatrial septum. Note the prominent visualization of the fossa ovalis (arrowhead), which is a common finding, as lipomatous hypertrophy typically spares the fossa ovalis. This could be misinterpreted as an atrial septal defect if the lipomatous hypertrophy is not noted. In addition, an atrial thrombus (arrow) is noted (see also image below). A myxoma would be in the differential diagnosis, but a myxoma classically arises from the interatrial septum in the region of the fossa ovalis.

Pericardium

Pericardial disease is evaluated initially with echocardiography. The most common reason for imaging is to assess effusion and tamponade. Evaluation of the pericardium with echocardiography is limited if no effusion exists or if effusion is complex. Echocardiography can cause misdiagnosis of pericardial cysts, tumors, and diaphragmatic hernias as effusion. MRI does not have these limitations.[26]

At MRI, simple axial and coronal or sagittal imaging planes are used. Effusions complicated with adhesions or loculations are clearly shown. Transudative effusion with low signal intensity on spin-echo images can be differentiated from exudative or hemorrhagic effusions that have high signal intensity. In addition, cine images can depict diastolic collapse of the chambers, which indicates tamponade. Pericardial thickening can be evaluated reliably with MRI, unlike with echocardiography. Pericardial thickness greater than 2 mm (typical) may suggest an inflammatory process in association with effusion.

MRI has high sensitivity in the diagnosis of constrictive pericarditis (see image below). The presence of pericardial thickening differentiates constrictive pericarditis from restrictive cardiomyopathy. Proper diagnosis is crucial; although presentations can be identical, treatments differ markedly. Constrictive pericarditis requires pericardiectomy; restrictive cardiomyopathy requires medical management. As a finding in constrictive pericarditis, diffuse pericardial thickening of 4 mm or greater has 93% accuracy.

Constrictive pericarditis: Steady-state gradient e Constrictive pericarditis: Steady-state gradient echo (SSFP [steady-state free precession]) image demonstrates thickening of the pericardium (arrow), with a resulting small right ventricular chamber. A pericardial effusion is also noted.

Other findings, such as atrial dilation and ventricular or septal alterations, can be present, but these are less specific. Inability to image pericardial calcification is a limitation in diagnosis because the specificity of pericardial calcification is high.

In the appropriate clinical setting, pericardial calcification on CT scans or on plain radiographs is diagnostic of constrictive pericarditis, and MRI is unnecessary. MRI is used to examine the minority of patients with symptoms of constrictive pericarditis in whom calcification is absent. MRI can also be used to examine asymptomatic patients and those with atypical symptoms in whom pericardial calcifications are found incidentally to ascertain whether constrictive pericarditis exists.

 

Future of Cardiac MRI

Artificial intelligence (AI) is revolutionizing the field of medical imaging—in particular, cardiovascular magnetic resonance (CMR)—by providing deep learning solutions for image acquisition, reconstruction, and analysis. Methods have been developed to enhance and expedite CMR data acquisition, image reconstruction, postprocessing, and analysis, and promising AI-based biomarkers have been identified for a wide spectrum of cardiac conditions.[29]

Nelson et al discussed the current role and future direction of cardiac MRI in refining diagnosis and personalizing management of ventricular arrhythmias. The capability of cardiac MRI with gadolinium contrast delayed-enhancement patterns and T1 mapping to determine the etiology of heart failure has been well established. Cardiac MRI can contribute to risk stratification, with precise and reproducible assessment of ejection fraction, quantification of scar and “border zone” volumes, and other indices. Detailed tissue characterization has begun to enable creation of personalized computer models to predict an individual patient's arrhythmia risk. When patients require ventricular tachycardia ablation, a substrate-based approach is frequently employed, as hemodynamic instability may limit electrophysiologic activation mapping. Beyond accurate localization of substrate, cardiac MRI could be used to predict the location of re-entrant circuits within the scar to guide ablation.[30]

One of the challenges of cardiac MRI involves compensation of respiratory motion, which causes the heart and surrounding tissues to move. Commonly used methods to overcome this effect—breath-holding and MR navigation—present shortcomings in terms of available acquisition time and need to periodically interrupt the acquisition, respectively. Santini et al presented a system of respiratory motion compensation to obtain information from abdominal ultrasound while continuously adapting the imaged slice position in real time. The technique enabled efficient imaging of the heart with resolution that would not be feasible with a single breath-hold.[31]