Real-Time Imaging of Cardiac Strain Using a FastHARP Pulse SequenceSmita Sampath, Andrew Derbyshire, Dara Kraitchman, Ergin Atalar, Nael Osman and Jerry L. Prince
Coronary artery disease is a leading cause of death in the
United States. Blockages in the coronary artery can, under stress, result
in the development of ischemic regions in the myocardium -- those with a
reduced flow of blood and oxygen. If this condition persists, these regions
infarcted -- tissue death or necrosis due to loss of blood flow. Depending
on the extent of the infarction, a cardiac arrest might occur. Early diagnosis
of coronary artery diseases can help prevent heart attacks. Stress tests
are normally conducted and the performance of the heart function is monitored
using an imaging modality. Since the mechanical and functional behavior of
ischemic and infarcted tissue is different from normal tissue, cardiologists
hope to use advanced imaging technology that would detect the onset of an
ischemic event during a stress test and quantify the extent of diseased (ischemic
or infarcted) tissue.
Several imaging modalities such as echocardiography, nuclear cardiology and
magnetic resonance imaging are currently used in the diagnosis of ischemic
heart disease. The echocardiographic stress test is clinically the most commonly
used method due to its cost-effectiveness and portability. It suffers, however,
from sub-optimal visualization of the cardiac segments and operator dependency
in the interpretation of its images. While nuclear cardiology is useful in
imaging the flow of blood supply in the myocardium, it has been less popular
due to the use of ionizing radiation. Magnetic resonance imaging (MRI) has
been gaining recognition as a safe, non-invasive, and superior imaging technique,
which has the potential to become a versatile cardiac imaging modality.
MRI can be used in the visualization of detailed cardiac anatomy, providing an accurate determination of regional wall thickness and muscle mass in the left ventricle. These are used to compute global health indices such as ejection fraction. MRI can also be used in myocardial perfusion studies to diagnose a reduced supply of blood to the myocardium. Both these techniques are incapable of characterizing the complicated motion within the myocardium itself. To achieve this cardiac motion analysis using magnetic resonance tagging, phase contrast MRI, and stimulated echo MRI (also known as DENSE) techniques were introduced.
Our research is based on the tagging
method which has proven to be sensitive in determining the motion of material
depicting cardiac strain. These strain maps are very useful in illustrating
localized healthy and diseased tissue regions and have great diagnostic potential.
Unfortunately, until recently, slow image acquisition times, and non-automated
and cumbersome image analysis methods limited the clinical use of MR tagging.
The harmonic phase (HARP) method, however, addresses these drawbacks
by providing an efficient fast imaging solution and automatic post-processing
methods for strain analysis. In the Fourier domain,
spatial modulation of magnetization (SPAMM) tagged MR data has several spectral
peaks. HARP uses the fact that the phase of the image corresponding to an off-origin
spectral peak is directly related to the tissue motion. Spatial derivatives
of harmonic phase are computed at each pixel to obtain strain measures.
computations are fast and automated, and provide displacement or strain maps
from the acquired tagged data in seconds. The post-processing tools have already
been developed and demonstrated in previous research.
Until recently, conventional MR tagging imaging techniques
were used to acquire the image data for HARP post-processing. These techniques
large (256 times 128-size) regions in Fourier space requiring long breath-holds.
For example, to measure cardiac motion in one slice and in any one direction,
it may take a breath-hold lasting 8–20 heartbeats to acquire a
movie sequence of tagged images. Two such long breath-holds are needed to
compute 2-D motion. Thus, for multiple slices, several long breath-holds
are required. These long breath-holds are inconvenient for patients. Also,
importantly, long imaging times result in a delay in the detection of any
ischemic event, if it were to occur. This is not ideal in a clinical stress
test environment as prolonged ischemia under stress (stress is drug administered
in an MR stress experiment) can lead to more serious complications. Thus
the need to develop faster HARP imaging techniques is crucial to patient
safety and convenience.
Here, we present a pulse sequence called FastHARP that exploits the fact that only two spectral peaks are sufficient for the computation of 2D-motion information. The FastHARP pulse sequence acquires multi-frame HARP images in~ a single heartbeat for a given tag orientation. Data acquired in only two heartbeats is used to compute in-plane quantities describing myocardial deformation for a single slice, thereby reducing the required breath-hold interval.
All experiments were conducted on a 1.5T Signa CV/i whole body magnetic resonance
system (GE Medical Systems, Waukesha, WI) equipped with 40 mT/m imaging
gradients with slew rates up to 150mT/m-ms. A gated, multi-phase, interleaved,
gradient-echo EPI pulse program was modified to provide a real-time HARP
imaging pulse sequence, FastHARP. A complete FastHARP acquisition requires
two heartbeats. A sequence timing diagram (Fig. 1) shows the FastHARP
pulse sequence over one heartbeat.
During the first heartbeat, vertical 1-1 SPAMM tags are generated at end-diastole
triggered by the R wave of the ECG by applying two 90 RF pulses (A and C),
separated by a tagging gradient (B) in the x direction. Crusher gradient
pulses are then applied to spoil residual magnetization in the transverse
plane. A tag separation of 8 mm is typically used.
Once the tags are created, acquisition of a small region in k-space around
the selected spectral peak begins. RF imaging pulses (D) with an incrementing
train of flip angles that bear a specific relation with each other are applied
to equalize the signal strength (magnitude) in each cardiac phase by compensating
for the tag fading caused by longitudinal relaxation and~
imaging pulses. To maximize the signal strength per cardiac phase, a program
is run before the scan to determine the optimal final flip angle. The choice
of the myocardial tissue T1 value used in the computation of the HARP image
contrast is important. It can be
shown that overestimation (resp. underestimation) of the actual T1 value
can result in dimming (resp. brightening) in the magnitude in each subsequent
TR acquisition. We used T1 = 850 ms in our computations since this yielded
images with visually uniform signal strength.
To shift the center of the acquisition window to the center of the harmonic
peak, the read dephaser gradient (E) and the phase encoding table (F) areas
are modified (Fig. 2). A gradient-echo EPI sequence (G) having 4 interleaved
shots is then used to acquire a 32 times 32 matrix in k-space centered on
the spectral peak. A bottom-up scheme is used to traverse the k-space during
with an echo-train of length 8. A receiver bandwidth of 62.5kHz was found
to be an optimal compromise between acquisition time and SNR for 32 readout
samples for the given slew rate. These parameters result in a TR of 9.7 ms.
The effective TE used was 2ms.
Depending on the heart rate, 7–20 images can be acquired
in one heartbeat, with a 40 ms total acquisition time for each image. The
is used to acquire HARP images having an orthogonal tag orientation. To accomplish
this, the frequency (Gx) and phase (Gy) directions are swapped in the second
We refer to the two-heartbeat FastHARP acquisition as the single-shot mode. FastHARP can also be run in a continuous mode in which HARP images are continuously acquired with tags alternating between vertical and horizontal orientations in successive heartbeats. In this mode, HARP strain (or displacement) images can be obtained every heartbeat by processing the images acquired during the preceding two heartbeats.
FIG. 1. A Timing Diagram describing the FastHARP Pulse Sequence
More details regarding the methods and analysis of the FastHARP data can be found in our publications listed below. The following section showcases some of our key results.
Results from Normal Volunteer Experiments
Experiments were conducted on human volunteers to visualize normal heart
motion. Figure 2 shows Eulerian strain maps overlaid by synthetic tags,
computed from the harmonic phase images of one volunteer. The time series
early systole (38.8 ms) and ends in diastole (562.6 ms). The temporal sampling
rate has been doubled by using a view-sharing type scheme to obtain one
image every 19.4ms. Observe the progressive increase in the
that can be visualized by the bending of the tag lines and the increased
dark blue coloration in the strain maps. The diastolic images show a gradual
decrease in the circumferential shortening.
FIG. 2. Eulerian strain maps with an overlay of synthetic tagged lines depicting the motion over one cardiac cycle. Strain images were computed from data obtained using the FastHARP sequence in a two heart-beat breath-hold from a normal volunteer. HARP post-processing tools were used to analyze the acquired images.
FIG. 3. (a) Avi movie of a synthetically tagged beating heart with selected points tracked in time. (b) Eulerian strain with a synthetic tag overlay. These movies were created from data acquired over a 2 heart-beat breath-hold using the FastHARP pulse sequence. To play, right click on the icon and play using Real Player.
During cardiac stress exams, a continuous monitoring of the cardiac function while the patient is breathing is most desirable. While the FastHARP pulse sequence can be used in a continuous monitoring mode, a comparison in the measures of myocardial function during free breathing and breath-hold studies is essential. The results below illustrate qualitative comparisons in Lagrangian strain measures in one normal volunteer and quantitative comparisons performed on six normal volunteers using the FastHARP pulse sequence.
FIG. 4. Representative Lagrangian endocardial strain curves for twelve segments in the myocardium using a two-heartbeat breath-held FastHARP acquisition ( dot-dashed black line) and a free-breathing two-heartbeat FastHARP acquisition (red solid line).
FIG. 5. (a) Correlation between Lagrangian myocardial circumferential strains between Y1 and Y2. (b) Bland-Altman plots for the comparison of Lagrangian circumferential strains between Y1 and Y2. Y1 stands for breath-held FastHARP, Y2 stands for non-breath-held FastHARP.
Note that the trends of both the curves in Figure 4 appear to be very similar. The correlation coefficients and the slopes in the linear regression plots for data obtained in Fig. 5(a) were: 1) 0.86 and 0.92 respectively between Y1 and Y2 in all six volunteers. Fig. 5(b) shows Bland Altman plots with mean differences of: –0.005 +/-0.037 (mean+/-standard deviation) between Y1 and Y2. We believe that the results are encouraging and reasonable for the intended use of this sequence to provide good qualitative strain measures in a cardiac stress-test protocol.
FastHARP with CSPAMM
Since the total flip angle of the 1-1 SPAMM rf pulses is 180 degrees, there is no dc peak immediately after tagging. Due to the longitudinal relaxation, however, there is a re-growth of the DC spectral peak and a corresponding decay of the harmonic peaks. The spectral contamination due to the dc interference effects results in an inaccurate quantification of deformation and an increased presence of artifacts in the resulting strain maps generated. It has been shown that these effects can be eliminated or at the least reduced using the CSPAMM imaging technique which is based on the subtraction of two images with complementarily signed tagging modulations. In this section, results from an implementation of a fast pulse sequence that acquires HARP images using CSPAMM in real-time is presented. In this mode, four heartbeats (instead of two) are required to generate a sequence of HARP strain images for each slice. The third and the fourth heartbeats are used to acquire harmonic peaks with a change in the sign of the second tagging rf flip angle. The first and second images are then subtracted from the third and fourth respectively to obtain CSPAMM corrected images. In the clinical monitoring mode, an update in myocardial strain is available every heartbeat by combining data from the last four subsequent heartbeats, resulting in real-time visualization of cardiac motion with better quality strain maps.
FIG. 6. Comparison of the Lagrangian strains obtained in one volunteer using two-heartbeat FastHARP sequence (red dot-dashed line) and a four-heartbeat FastHARP sequence with CSPAMM ( green solid line). Note the smoother and more reliable curves with CSPAMM, especially at later time frames.
Ischemic Dog Experiments with FastHARP [Kraitchman et. al.]
Experiments were performed on dogs to investigate the use of Fast Harmonic Phase (FastHARP) MRI for the quantitative, operator-independent detection of the onset of ischemia during acute coronary occlusion. The movie below illustrates end-systolic strain maps obtained during one of these experiments. Prior to the imaging, a balloon is inserted into the coronary artery. A continuous FastHARP acquisition for about a minute and 20s then follows. About 20s into the scanning, the balloon is inflated and ischemia is induced. At about 66s, the balloon is deflated. We found during this experiment that the strain maps indicated an abnormality in the vicinity of the ischemic region within a couple of seconds of the balloon inflation. This was much before any minor EKG changes could be noticed. Parts of the movie occur during free-breathing which is apparent by the jerky nature of the movie during those intervals. Note the recovery of the ischemic region after the balloon deflation.
1. Real-Time Imaging of Two-Dimensional Cardiac Strain Using a Fast HARP Pulse Sequence. Sampath S., Derbyshire A., Osman N.F., Atalar E., Prince J.L, Magnetic Resonance in Medicine, vol. 50, no. 1, pp. 154-163, July 2003.
2. Quantitative Ischemia Detection During Cardiac Magnetic Resonance Stress Testing Using Real-Time FastHARP. Kraitchman D.L., Sampath S., Castillo E., Derbyshire A., Bluemke D.A., Gerber B., Prince J.L., Osman N.F, Circulation, vol. 107, no. 15, pp. 2025-2030, 22 April 2003.
3. Real-time imaging of cardiac strain using FastHARP: a comparison between breath-hold and non-breath-hold studies. Sampath S., Derbyshire J.A., Osman N.F., Prince J.L. Proceedings of the Society for Cardiovascular Magnetic Resonance, 2003.
4. Real-Time Imaging of Myocardial Strain Patterns using a FastHARP sequence with CSPAMM. Sampath S., Derbyshire J.A., Osman N. F., Prince J.L. Proceedings of International Society of Magnetic Resonance in Medicine, 2002
5. Real-time Imaging of Cardiac Strain Using Fast HARP sequence. Sampath S., Derbyshire J.A., Osman N.F., Atalar E., Prince J.L. Proceedings of the International Society of Magnetic Resonance in Medicine, 2001.
6. Detetecting the Onset of Ischemia Using Real-Time HARP. Kraitchman D., Sampath S., Derbyshire J.A., Heldman A.W., Prince J.L., Osman N.F. Proceedings of the International Society of Magnetic Resonance in Medicine, 2001.
7. Synthetic tagged MR images for real-time HARP imaging. Osman N.F., Sampath S., Derbyshire J.A., Atalar E., Prince J.L. Proceedings of the International Society of Magnetic Resonance in Medicine, 2001.
8. Phantom Validation of the Fast HARP Pulse Sequence. Sampath S., Parthasarathy V., Prince J.L. IEEE International Symposium of Biomedical Imaging, 2002.
9. Real-Time Myocardial Tagging with Harmonic Phase MR Imaging: A Validation Study in Humans. Castillo E., Sampath S., Derbyshire J.A., Prince J.L., Osman N.F., Bluemke D.A. Proceedings of the Radiological Society of North America, 2002.