| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Technical Developments |
1 From the Cardiovascular Magnetic Resonance Unit, Royal Brompton Hospital, Sydney Street, London SW3 6NP, England. Received February 26, 2004; revision requested May 5; revision received May 20; accepted June 28. A.G.E. supported by grants from CORDA The Heart Charity. J.C.M. supported by the British Heart Foundation. Address correspondence to D.J.P. (e-mail: d.pennell@imperial.ac.uk).
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMaterials and Methods ResultsDiscussionREFERENCES |
|---|
© RSNA, 2005
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMaterials and Methods ResultsDiscussionREFERENCES |
|---|
To our knowledge, the sequence of choice for perfusion cardiovascular MR imaging has not yet been determined. The majority of perfusion cardiovascular MR studies have used fast single-shot gradient-echo (FLASH) (6–9) and hybrid echo-planar imaging (EPI) (3,10–12) sequences. More recently, steady-state free precession sequences have also been used (13), but experience with these is limited. FLASH is the most established sequence, and it is robust in practice; however, its spatiotemporal resolution (approximately 180 msec per image at approximately 3 mm) limits the number of sections that can be acquired during each cardiac cycle, particularly during stress with induced tachycardia. This is a potentially important limitation, because comprehensive myocardial coverage is crucial if perfusion cardiovascular MR is to become clinically relevant in the diagnosis of CAD. In addition, the short time between radiofrequency pulses and the spoiling of transverse magnetization means that the flip angle used must be low, which limits the potential contrast-to-noise ratio (CNR) of the sequence. The EPI sequence (11) has advantages over the FLASH sequence because it collects multiple (typically four) data lines after each radiofrequency pulse. The increased repetition time permits higher flip angles for potentially higher CNR values and faster imaging for comprehensive ventricular coverage with good spatiotemporal resolution. Parallel imaging and partial Fourier methods may still improve the performance of both sequences.
To our knowledge, there has been no previously published direct comparison of FLASH and EPI techniques to determine their relative merits in patients with CAD. We therefore conducted our study to compare FLASH and EPI MR techniques regarding the relative CNR, spatiotemporal resolution, size of inducible perfusion defects, and presence of artifact in patients with CAD.
|
Materials and Methods |
|---|
|
TOPABSTRACTINTRODUCTIONMaterials and Methods ResultsDiscussionREFERENCES |
|---|
Imaging
All studies were performed with a 1.5-T imager (Sonata; Siemens, Erlangen, Germany) and a four-element (two anterior and two posterior coil elements) phased-array cardiac coil. The study was approved by the local ethics committee, and each subject gave written informed consent. Each subject abstained from caffeine or other adenosine antagonists for 24 hours prior to each study. The patients were split into two groups. Within each group, each subject underwent a rest and adenosine stress first-pass myocardial perfusion cardiovascular MR study with EPI; on a separate day, each subject underwent a study with the FLASH sequence. The EPI study was performed first in each patient.
Group 1 was composed of nine subjects (seven men and two women). Three had single-vessel disease, one had two-vessel disease, and five had triple-vessel disease. Four of the volunteers had a known previous myocardial infarction. In this group, the EPI sequence was set to have a spatial resolution similar to that of the comparative FLASH sequence, which resulted in the EPI sequence having a faster temporal resolution (repetition time msec/echo time msec, 5.6/1.2; read field of view [FOV], 34–40 cm; phase FOV, 75.0%–87.5% of the read FOV; base resolution, 128; pixel size ranged from 2.7 x 3.6 mm to 3.1 x 4.1 mm; nonselective saturation pulse, 90°; flip angle, 30°; EPI factor, 4; 58 msec from the saturation pulse to the center of k space acquired by the 10th radiofrequency pulse halfway through the 101 msec image time, total time per section, 109 msec [for 75% phase FOV]; section thickness, 8 mm; bandwidth, 1860 Hz/pixel) than the FLASH sequence (2.0/1.0; read FOV, 34–40 cm; phase FOV, 68.8% of the read FOV; base resolution, 128; read pixel size, 2.7 x 2.9 mm; pixel size, 3.1 x 3.3 mm; nonselective saturation pulse, 90°; flip angle, 8°; 90 msec from saturation to the center of k space acquired by the 42nd radiofrequency pulse halfway through the 163 msec image time; total time per section, 173 msec; section thickness, 8 mm; bandwidth, 780 Hz/pixel). No filtering was applied to the reconstruction of either sequence.
Group 2 was composed of six men. Two subjects had single-vessel disease, two had two-vessel disease, and two had triple-vessel disease. In this group, the EPI sequence parameters were altered to make the temporal resolution similar to that of the FLASH sequence, and this allowed improved spatial resolution for the EPI sequence (6.2/1.3; read FOV, 34–40 cm; phase FOV, 75.0%–87.5% of the read FOV; base resolution, 160; pixel size ranged from 2.1 x 2.1 mm to 2.5 x 2.5 mm; flip angle, 30°; 101 msec from saturation to the center of k space acquired by the 16th radiofrequency pulse halfway through the 186 msec image time; total time per section, 195 msec; bandwidth, 1838 Hz/pixel). The FLASH parameters were kept the same for all patients in both groups.
Each 0.1-mmol per kilogram of body weight gadolinium bolus (Magnevist; Schering, Berlin, Germany) was injected via an 18-gauge cannula in the right antecubital fossa at a rate of 7 mL/sec by using a Spectris power injector (Medrad, Indianola, Pa) with a 10-mL normal saline flush. The flush volume was sufficient to ensure complete delivery of gadolinium into the patient. The patients were asked to hold their breath in end expiration during each first-pass study for as long as was comfortable. An adenosine (140 µg/[kg · min]) stress study was performed at least 20 minutes after the rest study. Adenosine was infused for 4 minutes via a 20-gauge cannula in the arm opposite to that in which the gadolinium was given. Heart rate was continuously monitored throughout each study. Blood pressure was measured at 3-minute intervals during the rest studies and at 1-minute intervals during adenosine infusion. For each study, three short-axis sections were acquired for each cardiac cycle over the course of 50 cardiac cycles. In each study, the basal section was positioned to avoid the left ventricular outflow tract; the typical gap between each short-axis section was 6–10 mm, and it was adjusted to the length of the left ventricular outflow tract. The apical short-axis section was acquired first, and the more basal sections were acquired subsequently; thus, the most basal short-axis section was acquired in diastole, and the occurrence of the left ventricular outflow tract descending into the image plane was reduced. Short-axis images obtained during the first visit were used to obtain short-axis images at the second visit, thus ensuring consistent imaging planes at the two visits.
Image Analysis
All image analysis was performed off line by using dedicated cardiovascular MR software (CMRtools; Cardiovascular Imaging Solutions, London, England). All images were randomized, and the scorers were blinded to the sequence used and the patient’s name. Four scorers (A.G.E., P.D.G., S.K.P., and T.M.C., with 3, 12, 3, and 3 years of experience, respectively) were used for image analysis.
Length of breath was determined (A.G.E.) by reviewing the findings of each first-pass perfusion study. In each study, CNR was measured by using a previously described technique (13). A region of interest (ROI) was drawn in each wall of the myocardium (anterior, lateral, inferior, and septum) in only the midventricular rest and stress short-axis section (T.M.C.) (Fig 1). ROI area ranged from 60 to 100 mm2. The analysis software package did not allow the rotation of the ROI to be adjusted for the four walls of the myocardium or transferred from study to study; therefore, a new ROI was drawn for each region analyzed. Thus, four CNR values were generated for each rest study, and four CNR values were generated for each stress study. Particular care was taken to avoid contamination of the ROI by signal from the right and left ventricular blood pools. ROI positions were adjusted manually to compensate for respiratory and cardiac motion, thus ensuring that the ROI remained in the same region of myocardium throughout the study. This was necessary in all patients for some images, especially at the end of the breath hold. ROI repositioning was primarily guided by the previous cardiac cycle image rather than by intramyocardial landmarks.
|
|
|
|
Results |
|---|
|
TOPABSTRACTINTRODUCTIONMaterials and Methods ResultsDiscussionREFERENCES |
|---|
|
|
|
|
|
Induced Perfusion Defects
Of the 30 rest-stress perfusion studies performed, one observer (A.G.E.) considered inducible perfusion defects present in 22 (12 EPI studies and 10 FLASH studies), and one observer (P.D.G.) considered inducible perfusion defects present in 23 (12 EPI studies and 11 FLASH studies). In the eight studies in which the first observer (A.G.E.) did not consider an inducible perfusion defect to be present, the second observer (P.D.G.) was in agreement, with the exception of one FLASH study, in which the second observer considered a small inducible perfusion defect to be present.
There was no significant difference in the area of induced perfusion defects with either sequence with the same spatial (EPI = 260 mm2± 300 vs FLASH = 300 mm2± 280; P = .39) and temporal (EPI = 440 mm2± 250 vs FLASH = 430 mm2± 510; P = .9) resolution. There was no significant difference in the maximal transmural extent of induced perfusion defects with either sequence with the same spatial (EPI = 35% ± 27% vs FLASH = 32% ± 26%; P = .33) and temporal (EPI = 73% ± 20% vs FLASH = 55% ± 44%; P = .16) resolution.
Infarctions
The five myocardial infarctions were all apparent on the first (rest) perfusion images; however, they were less apparent on the second (stress) perfusion images. This is due to residual gadolinium remaining in the infarct at the time of the stress study. The residual gadolinium reduces the relative hypoenhancement of the infarct in comparison to the normal myocardium at first pass of the gadolinium bolus.
|
Discussion |
|---|
|
TOPABSTRACTINTRODUCTIONMaterials and Methods ResultsDiscussionREFERENCES |
|---|
There was a wide range in the CNR values calculated with both sequences. A spread in CNR values is to be expected because of the range of distances that regions analyzed were from the surface coils; however, the reason why spread is greater with the EPI sequence is not known.
It is of note that with the higher-spatial-resolution EPI sequence, there was no deterioration in the CNR in comparison to the lower-spatial-resolution EPI sequence. Image signal intensity and, consequently, CNR is proportional to the voxel volume if all other imaging parameters are kept constant. Thus, with the higher spatial resolution, and, consequently, lower voxel size, the CNR decrease may have been expected for the EPI sequence. However, the higher-spatial-resolution EPI sequence had a longer saturation delay time compared with the lower-spatial-resolution EPI sequence (100 msec and 60 msec, respectively). The longer saturation delay and image acquisition time of the higher-spatial-resolution EPI sequence may have compensated for the decrease in voxel size.
The superiority in CNR with EPI was present in both the rest and the stress images. By performing the adenosine stress studies, however, we were able to make some assessment of the ability of the sequences to depict inducible perfusion defects. As there is almost always an increase in heart rate at stress, we thought it was probable that the peak concentration of gadolinium in the myocardium would be higher than that at rest, with a resultant increase in signal. With the greater signal, it was expected that the CNR in the myocardium would be enhanced at stress in comparison with that at rest; however, we only found significantly increased enhancement between rest and stress CNR in the high-temporal-resolution EPI studies.
A weakness of perfusion cardiovascular MR imaging has been the presence of a dark ringing artifact in the subendocardium, which is most apparent at maximum left ventricular blood pool brightness. This artifact is particularly troublesome because it may mimic the subendocardial perfusion abnormalities seen in patients with CAD. This artifact has often been ascribed to Gibbs ringing. Recent work, however, has indicated that this artifact may also be due to cardiac motion during image acquisition (14). With its potential for shorter imaging time, the EPI sequence would be less prone to this artifact.
The lower bandwidth used in the FLASH sequence was necessary to maintain signal-to-noise ratio, as only lower flip angles were possible with the short repetition time of this sequence. On the basis of simulations of the effect of different flip angles and repetition times on gadolinium-enhanced myocardium with an estimated T1 of 200 msec, the optimal flip angles were 25° for both the lower- and the higher-spatial-resolution EPI sequence and 13° for the FLASH sequence. The decrease in simulated myocardial signal-to-noise ratio with the EPI sequence and flip angle (30°) used in this study is 2%. The decrease in simulated myocardial signal-to-noise ratio with the FLASH sequence and flip angle (8°) used in this study is 18%. Although use of optimal flip angle would have reduced the difference in CNR between the EPI and FLASH sequences, it clearly would not account for the disparity in CNR between the EPI and FLASH sequences as measured in this study (46% and 51%, depending on sequence resolution).
In this study, a relatively high dose of gadolinium was used (0.1 mmol/kg) for both the rest and the stress studies. The sequences used in this study had relatively long saturation delay times (EPI sequence, 60 and 100 msec; FLASH sequence, 90 msec). With saturation delay times this long, an arterial input function would be clipped by using this dose of gadolinium; thus, it would be unsuitable for use in quantitative perfusion analysis. In this study, however, we were not performing quantitative perfusion analysis; instead, we aimed to optimize the signal in the myocardium, which is achieved with a high dose of gadolinium and long saturation recovery.
An important aspect of this study was that it was performed in a manner that affords nearly comprehensive (16 of 17 segments) segmental myocardial coverage (15), which demands a minimum of three sections. The sequence parameters compared in this study are all potentially appropriate for clinical use; while further sequence developments for perfusion cardiovascular MR are likely, this study approximates a clinically viable scenario. There are several areas of new development in perfusion cardiovascular MR imaging, including parallel imaging (16) and unaliasing by Fourier encoding the overlaps by using the temporal dimension (commonly referred to as the UNFOLD method) (17), both of which improve the temporal resolution of the sequences to which they are applied. This study did not compare sequences by using these new techniques; however, these techniques may prove to be important in future perfusion cardiovascular MR applications, and further comparative studies would be helpful.
Rest studies were performed first. Although at least a 20-minute interval was allowed prior to the stress study, it is possible that some interference with the stress study could have occurred; however, the blood and uninfarcted myocardium gadolinium clearance is largely complete by this time.
The EPI study was always performed first. Randomization of the order of sequences would have removed this potential distorting factor. There was, however, no significant difference in the hemodynamic recordings between the studies.
CNR was calculated from the midventricular short-axis section; only this section usually provided the clearest views of the myocardium, without the potential influences of the left ventricular outflow tract and partial voluming found with the basal and apical short-axis sections, respectively. It was reasoned that the comparison of CNR values for the midventricular short-axis section would be similar to that of the other short-axis sections.
The CNR of the inducible perfusion defects was not measured in this study. Because of the inevitable heterogeneous nature of signal within induced perfusion defects, obtaining reliable measurements of CNR in these defects would be difficult.
This was a feasibility study that was not intended to show differences in diagnostic accuracy between sequences; for this purpose, a much larger study would be required. It is likely, however, that the technique with the best imaging characteristics would provide superior accuracy.
In this study, an EPI sequence had CNR and spatiotemporal resolution superior to that of a FLASH sequence for perfusion cardiovascular MR imaging. There was, however, no significant difference in artifact or size and transmural extent of induced perfusion defects between the two sequences. Further comparison is warranted to determine whether there is a difference in diagnostic accuracy between the two sequences.
|
FOOTNOTES |
|---|
D.J.P. is a consultant to Siemens, has served on an advisory board for Nycomed Amersham, and is a director of Cardiovascular Imaging Solutions.
Author contributions: Guarantor of integrity of entire study, D.J.P.; study concepts, A.G.E., P.D.G., D.N.F., D.J.P.; study design, A.G.E., P.D.G., S.K.P., J.C.M., D.J.P.; literature research, A.G.E., P.D.G.; clinical studies, A.G.E., P.D.G., T.M.C., J.C.M., S.K.P.; experimental studies, A.G.E., P.D.G.; data acquisition, A.G.E., P.D.G., T.M.C., S.K.P., J.C.M.; data analysis/interpretation, A.G.E., P.D.G., T.M.C., D.N.F., D.J.P.; statistical analysis, A.G.E., D.J.P.; manuscript preparation, A.G.E., P.D.G., T.M.C.; manuscript definition of intellectual content, A.G.E., P.D.G., D.N.F., D.J.P., J.C.M., S.K.P.; manuscript editing, A.G.E., P.D.G., D.N.F., D.J.P.; manuscript revision/review, A.G.E., P.D.G., S.K.P., D.N.F., D.J.P.; manuscript final version approval, all authors
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMaterials and Methods ResultsDiscussionREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  A. B. Wolbarst and W. R. Hendee Evolving and Experimental Technologies in Medical Imaging Radiology, January 1, 2006; 238(1): 16 - 39. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
= CNR values from rest studies. 
= CNR values from stress studies. CNR values were calculated from only the middle short-axis section in each study.