| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Technical Developments |
1 From the Department of Medicine, Division of Cardiology (J.G., J.A.C.L., B.L.G., K.C.W.), Department of Electrical Engineering (S.S., J.L.P., N.F.O.), and Department of Radiology (D.A.B., N.F.O.), Johns Hopkins University, Baltimore, Md. Received October 15, 2003; revision requested January 8, 2004; revision received January 12; accepted February 2. Supported by NIH/NHLBI grants R29HL-47405, HL-4590, and N01-HC95162. J.G. supported by a Boehringer-Ingelheim grant of the Fédération Française de Cardiologie (Reims, France). Address correspondence to J.G., Henri Mondor University Hospital, Cardiologie 8ème, 51 Avenue du Maréchal de Lattre de Tassigny, 94010 Créteil, France (e-mail: jgarot@free.fr).
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMaterials and Methods Results DiscussionAppendix REFERENCES |
|---|
© RSNA, 2004
Index terms: Heart, MR, 511.121412, 511.12143 • Magnetic resonance (MR), comparative studies • Magnetic resonance (MR), contrast enhancement, 511.12143 • Myocardium, infarction, 511.814
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMaterials and Methods Results DiscussionAppendix REFERENCES |
|---|
In contrast to conventional tagging, strain-encoded MR imaging is a technique that uses tag surfaces that are parallel, not orthogonal, to the image plane, combined with out-of-plane phase-encoding gradients in the perpendicular section-select direction (14). Because local frequency of the tag pattern is related to myocardial strain (15,16), we hypothesized that strain-encoded MR imaging might provide direct myocardial longitudinal strain (Ell) imaging embedded on short-axis images of the LV. This technique may have potential in the online assessment of intrinsic myocardial contractility and may therefore represent a powerful addition to the assessment of myocardial viability by quantifying local function automatically. Thus, the purpose of our study was to prospectively evaluate strain-encoded MR imaging as a method for direct imaging of regional LV function that precludes the need for postprocessing and can be used in combination with contrast material–enhanced MR imaging in the assessment of local myocardial viability.
|
Materials and Methods |
|---|
|
TOPABSTRACTINTRODUCTIONMaterials and Methods Results DiscussionAppendix REFERENCES |
|---|
Data Acquisition
Clinical research studies were performed with a 1.5-T MR imager (Signa; GE Medical Systems, Milwaukee, Wis). Anterior and posterior phased-array surface coils were used for signal acquisition. Strain-encoded MR images and 3D tagged MR images, which were used as a reference for measurement of myocardial strain, were acquired sequentially in all subjects. In addition, gadolinium (Gd) diethylenetriaminepentaacetic acid (DTPA)– enhanced MR imaging was performed in patients with MI who underwent cardiac MR imaging 3 days ± 1 after acute MI.
Strain-encoded MR imaging.—Strain-encoded MR imaging uses tag planes parallel and not orthogonal to the LV short-axis image plane. We used a modified 1–1 spatial modulation of magnetization tagging pulse sequence with a magnetic field gradient oriented in the section direction to spatially modulate the longitudinal magnetization in a sinusoidal pattern, so that planes of constant sinusoidal phase are parallel to the image plane. During systole, the healthy human LV undergoes displacement and deformation along its long axis (longitudinal direction). The tag pattern moves with the tissue and undergoes compression, while the LV shortens along the long axis. Thus, myocardial tissue strain affects the frequency of the tag pattern locally. More specifically, myocardial Ell is proportional to the local frequency vector of the tag pattern in the longitudinal (z) direction.
Strain-encoded MR imaging is used to measure the frequency and orientation of the tag pattern within each voxel and provides all that is needed to calculate a high-resolution map of Eulerian Ell on a short-axis section (14). Strain is defined as the change in length per unit length. To measure the local frequency within each voxel, we used a modified electrocardiographically triggered segmented k-space fast gradient-echo imaging pulse sequence (17). From the long-axis scout image, we prescribed three equidistant sections that spanned the entire LV. Out-of-plane 1–1 SPAMM tags with a tag period of 2.5 mm were applied at end diastole, parallel to the image plane, and a phase-encoding gradient (Gz)—which we call the tuning gradient—was applied in the z direction (section-select direction) before readout. For measurement of Ell, the method requires acquisition of two images per location corresponding to two different tuning gradients orthogonal to the image plane. The selected tuning values depend on initial tag separation and expected changes in tag frequency as a result of regional tissue shortening. Tag frequency corresponds mathematically to the reciprocal of tag separation. We chose a tag separation of 2.5 mm, which corresponds to a 0.40-mm–1 tag frequency. Since we expected maximal systolic longitudinal shortening to be about 15%, the corresponding tag frequency was 0.46 mm–1. These two tuning values (0.40 and 0.46 mm–1), chosen to correspond to a 0%–15% range of strain values (14), were applied at each section location (one per breath hold) (Fig 1). From these two images, a strain image was automatically produced, of which signal intensity is directly related to Ell (Appendix). Typical imaging parameters were a 400-mm field of view; 10-mm section thickness; 256 x 160 matrix; repetition time msec/echo time msec, 6.8/1.8; flip angle, 20°; one signal acquired; and 1.56 x 2.50-mm in-plane resolution. Breath-hold time was about 10 seconds. This sequence was used to generate five to seven images throughout systole with a temporal resolution of about 50 msec.
|
Delayed Gd-DTPA–enhanced MR imaging.—Gd-DTPA–enhanced MR imaging has been used to characterize acute MI and represents an accepted technique for infarct delineation in humans (9,10). Because of the increased volume of distribution and altered wash-in and washout kinetics of the extravascular contrast agent within infarcted myocardium, images acquired 10–15 minutes after contrast agent injection demonstrate regional hyperenhancement in the infarcted area relative to the remote area. After conventional 3D tagging and strain-encoded acquisitions were completed, images were acquired 10–15 minutes after administration of an intravenous bolus of 0.1 mmol/kg Gd-DTPA (Omniscan; Sanofi-Synthélabo, Paris, France). We prescribed six short-axis sections spanning the entire LV from base to apex. We used an inversion-recovery prepared fast spoiled gradient-echo pulse sequence with an inversion time between the 180° inversion-recovery pulse and the radiofrequency imaging pulse of 200–270 msec (250–270 msec in eight patients and 200 msec in one), which was found to provide optimal contrast between remote and infarcted myocardium. Imaging parameters were as follows: 256 x 192 image matrix, 6.0/2.3, 360-mm field of view, 7-mm section thickness, and 20° flip angle. Each section was acquired in a single breath hold of about 8–10 seconds. Mean total time spent in the imager for a patient was about 45 minutes.
Data Analysis
Strain-encoded MR imaging.—Two short-axis images with two different tuning values in the direction orthogonal to the image plane were acquired at each section location. From these images, a dense map of Ell was automatically generated by using a mathematic equation, as described in the Appendix. On this strain image, coordinates of the posterior right ventricular–LV insertion point were calculated on the most basal section and used as reference landmarks for segmentation of the LV into five segments: inferior, inferolateral, lateral, anterior, and septal wall. For each short-axis section of the heart, the same five-segment model was then applied for cross-registration between imaging modalities, and mean maximal Ell (five points per segment) was automatically computed at end systole in the subendocardium and subepicardium of each myocardial segment at the base, middle LV, and apex. Ell depends directly on the intensity of the Ell image and was automatically determined (Appendix). For measurement of maximal systolic strain, the still frame exhibiting maximal deformation at end systole was selected. For assessment of reproducibility, analysis was performed by two independent observers (J.G. and N.F.O., both with 9 years experience in cardiac MR imaging), who were blinded to each other and other data. Also, the area of dysfunctional myocardium was visually assessed on the strain image and manually contoured (J.G.).
Conventional 3D tagged MR imaging.—Maximum myocardial Ell was assessed (B.L.G., 9 years of experience with cardiac MR imaging) off-line at end-systole in the subendocardium and subepicardium by using an established tracking motion technique (13), as previously described (16). Images were processed by using a validated software program that requires interactive and time-consuming detection of myocardial contours and tag lines from short- and long-axis tagged cardiac images to generate a dense motion map (13). Ell was computed at the base, middle LV, and apex. The time required for a complete quantitative analysis was about 6–8 hours per subject. To ensure adequate cross-registration between strain-encoded, contrast-enhanced, and 3D tagged MR imaging, the three anatomic levels along the LV long axis were automatically copied and acquired with each imaging modality.
Delayed Gd-DTPA–enhanced MR imaging.—The same five-segment model was applied to identical section locations by using the same landmark for cross-registration. Segments were categorized as infarcted or not, on the basis of the presence or absence of hyperenhancement on delayed contrast-enhanced images (J.A.C.L., 15 years of experience with cardiac MR imaging). In each segment, infarct transmurality was determined visually as being more or less than 50% of myocardial wall thickness. For the purpose of analysis, infarctions were considered nontransmural when less than 50% of myocardial wall thickness and transmural when more than 50%. Myocardial segments that were immediately contiguous to the infarct region either radially or longitudinally were labeled as adjacent. Segments that were neither infarcted nor adjacent were categorized as remote. For each segment, the analysis was performed in the subendocardium (inner half) and subepicardium (outer half). Areas of hyperenhancement were contoured manually (J.A.C.L.).
Statistical Analysis
Values are shown as mean and standard error. Comparisons between myocardial strain obtained with strain-encoded MR imaging in remote, infarcted, and adjacent myocardium were assessed with repeated-measures analysis of variance with post-hoc Tukey correction. Comparisons between the two methods for assessment of Ell were performed with the paired Student t test. Correlations between the two techniques were assessed with linear regression analysis and Bland-Altman plots. Interobserver reproducibility of strain-encoded MR imaging was assessed with linear regression analysis, and the coefficient of variability between the two series of measurements was computed. Proportions were compared with 
2 analysis. All tests were two tailed and considered to indicate a statistically significant difference when the P value was less than .05.
|
Results |
|---|
|
TOPABSTRACTINTRODUCTIONMaterials and Methods Results DiscussionAppendix REFERENCES |
|---|
|
|
Strain-encoded MR Imaging versus Conventional 3D Tagged MR Imaging
Values of maximal systolic Ell obtained directly from strain-encoded MR images and standard 3D tagged MR images are shown in patients with infarction (Table). The concordance between Ell maps as assessed with strain-encoded and conventional 3D tagged MR imaging is illustrated in a patient with anterior MI in Figure 4. For pooled data in patients with infarcted, adjacent, or remote myocardium, there was a good correlation between the two methods for assessment of Ell (r = 0.75; standard error, 1.88; P < .01; Fig 5a). When comparing individual differences between the two techniques in each segment, we observed slightly lower Ell values with strain-encoded MR imaging at the higher range and slightly higher values at the lower range, as reflected by the Bland-Altman plot (Fig 5b). The time for complete analysis of Ell in a single subject was typically about 6–8 hours with 3D tagged MR imaging and less than 30 seconds with strain-encoded MR imaging. From the two phase-encoded images, a strain image is produced in a fraction of a second (see equations in the Appendix). The LV myocardium is then divided in equally distributed radial sectors, and quantitative data are obtained automatically in about 5 seconds (for time for segmentation of the LV and for calculation of myocardial strain, see Appendix).
|
|
|
|
|
Discussion |
|---|
|
TOPABSTRACTINTRODUCTIONMaterials and Methods Results DiscussionAppendix REFERENCES |
|---|
Importance of Myocardial Strain in Infarcted, Adjacent, and Remote Myocardium
By using MR tagging in sheep, Kramer et al (18) reported a decrease in circumferential and longitudinal shortening strain in adjacent myocardium relative to remote noninfarcted regions. In human patients, circumferential shortening strain was reduced in remote noninfarcted myocardium in comparison with that in human subjects (19). In dogs, Lima et al (20) reported impaired systolic wall thickening in the nonischemic myocardium immediately adjacent to the ischemic region and, to a lesser degree, in remote regions. In concordance with previous findings, we report a decrease in longitudinal shortening strain in adjacent myocardium after acute MI, as compared with that of remote myocardium. We also report a decrease in tissue strain in remote myocardium early after acute MI versus tissue strain in the same segments in control subjects. The regional dysfunction in adjacent and remote myocardium may be due to increased wall stress or mechanical tethering to infarcted regions and may be an important determinant of LV remodeling after MI (18–20). Moreover, by depicting subtle differences in regional function within ischemic, adjacent, and remote myocardium, strain-encoded MR imaging enables a detailed assessment of LV mechanics during LV postinfarct remodeling.
Advantages
The technique provides direct imaging and assessment of Ell and permits complete bypass of tedious interactive tracking of LV contours and tag motion over time from both short- and long-axis tagged images. In addition, tracking motion techniques with 3D tagged MR imaging require the use of complex algorithms to reconstruct a 3D strain tensor. A dense strain map is produced with interpolation and averaging. Fusing of short- and long-axis data requires careful registration, which may be altered by patient movement over a long series of breath holds. In contrast, strain-encoded MR imaging provides greater spatial resolution for pixel-by-pixel assessment of myocardial strain.
It has been shown that accurate information about tag motion and tissue deformation can be obtained from one spectral peak in k space for each direction of tag line (15,16). The development of a fast imaging pulse sequence with limited k-space acquisition may have potential for real-time strain-encoded MR imaging. This would be of great importance for real-time quantitative monitoring of regional LV function during cardiac stress MR imaging.
Limitations
This work does not provide new pathophysiologic insights, but its aim was to present initial validation of an attractive and direct strain imaging and measurement technique. Three sections were acquired per patient, but there is no technical limitation for the number of sections one can acquire with strain-encoded MR imaging. The complex mechanics of the LV are best described when myocardial strain can be assessed in various directions. If used in combination with harmonic-phase MR imaging (16), which provides detailed assessment of two-dimensional myocardial strain in the circumferential and radial directions, the automated detailed assessment of 3D cardiac mechanics can be implemented in humans. Maximal systolic strain was measured for validation purposes, but strain-encoded MR imaging enables determination of myocardial strain at each phase during systole, with a temporal resolution of about 50 msec. However, tag fading during diastole, along with 50-msec temporal resolution, did not allow for robust and accurate assessment of postsystolic thickening.
The two tuning values used in this study were predetermined on the basis of the normal range of Ell in healthy subjects. Comparison with 3D tagged MR imaging shows slightly lower strain values with strain-encoded MR imaging at the higher range and slightly higher values at the lower range. However, differences between the two modalities are moderate at the extreme range of values (absolute difference in longitudinal shortening strain is <4% in more than 95% of patients). This indicates that slight variations in tuning values have only moderate impact on strain measurement. The method requires two short breath holds per location, but the total acquisition time for assessment of Ell is substantially reduced, since long-axis sections of the heart are not needed. Although quantitative myocardial strain imaging may provide more information than visual assessment, the clinical relevance of the technique remains to be formally demonstrated.
Strain-encoded MR imaging provides direct spatially resolved imaging and quantification of myocardial Ell in humans without the need for additional time-consuming postprocessing. In patients with acute MI, we believe on-line quantification of regional LV contractility may represent an important addition to the evaluation of ischemic myocardial damage with contrast-enhanced MR imaging. We also believe this integrated approach has great potential for improved comprehensive evaluation of local myocardial viability.
|
Appendix |
|---|
|
TOPABSTRACTINTRODUCTIONMaterials and Methods Results DiscussionAppendix REFERENCES |
|---|
) is applied in the section-select or z direction, which corresponds to a tuning frequency in the z direction by using the following equation:
|
The resulting image (I) is the integral in the z direction of the longitudinal magnetization (M), multiplied by the tuning frequency factor over the section profile (s), as follows:
|
Two images are acquired: IL(x,y,t) = I(x,y,t;
L) and IH(x,y,t) = I(x,y,t;H), where L and H are the low- and high-tuning frequencies, respectively.
Strain Image
The local frequency vector is µ(x,y,t), which is the component in the z direction that depends on the through-plane strain, and it can be computed with the following equation:
|
The local strain can then be computed with the following equation:
|
The computed strain is then used to color the image of the heart. Blue indicates no contraction, while red indicates contraction.
|
ACKNOWLEDGMENTS |
|---|
|
FOOTNOTES |
|---|
Author contributions: Guarantors of integrity of entire study, J.G., N.F.O.; study concepts, J.G., J.A.C.L., B.L.G., J.L.P., N.F.O.; study design, J.G., J.A.C.L., J.L.P., N.F.O.; literature research, J.G., N.F.O., J.A.C.L.; clinical studies, J.G., J.A.C.L., B.L.G., K.C.W., D.B.; experimental studies, J.G., S.S., J.L.P., N.F.O.; data acquisition, J.G., J.A.C.L., B.L.G., K.C.W., D.B., N.F.O.; data analysis/interpretation, J.G., B.L.G., K.C.W., N.F.O., S.S.; statistical analysis, J.G., B.L.G., S.S.; manuscript preparation, J.G., J.A.C.L., N.F.O., J.L.P., K.C.W.; manuscript definition of intellectual content, J.G., J.A.C.L., N.F.O., J.L.P.; manuscript editing, J.G., N.F.O.; manuscript revision/review, J.G., J.A.C.L., J.L.P., N.F.O., B.L.G., S.S.; manuscript final version approval, all authors
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMaterials and Methods Results DiscussionAppendix REFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  M. Neizel, D. Lossnitzer, G. Korosoglou, T. Schaufele, H. Peykarjou, H. Steen, C. Ocklenburg, E. Giannitsis, H. A. Katus, and N. F. Osman Strain-Encoded MRI for Evaluation of Left Ventricular Function and Transmurality in Acute Myocardial Infarction Circ Cardiovasc Imaging, March 1, 2009; 2(2): 116 - 122. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 T. P. Abraham, V. L. Dimaano, and H.-Y. Liang Role of Tissue Doppler and Strain Echocardiography in Current Clinical Practice Circulation, November 27, 2007; 116(22): 2597 - 2609. [Full Text] [PDF]
|
|
|
|
|
|
A. Schmidt, C. F. Azevedo, A. Cheng, S. N. Gupta, D. A. Bluemke, T. K. Foo, G. Gerstenblith, R. G. Weiss, E. Marban, G. F. Tomaselli, et al. Infarct Tissue Heterogeneity by Magnetic Resonance Imaging Identifies Enhanced Cardiac Arrhythmia Susceptibility in Patients With Left Ventricular Dysfunction Circulation, April 17, 2007; 115(15): 2006 - 2014. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 H. Ashikaga, S. R. Mickelsen, D. B. Ennis, I. Rodriguez, P. Kellman, H. Wen, and E. R. McVeigh Electromechanical analysis of infarct border zone in chronic myocardial infarction Am J Physiol Heart Circ Physiol, September 1, 2005; 289(3): H1099 - H1105. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
= infarcted myocardium. 
= adjacent myocardium. 
= remote myocardium.