| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Cardiac Imaging |
1 From the Departments of Cardiology (Y.C., P.L., J.E.W.) and Magnetic Resonance Imaging (C.T., F.G., A.L., O.R., S.R., P.M.W., F.B.), Centre Hospitalier Universitaire, 10, Boulevard Maréchal de Lattre de Tassigny, 21034 Dijon, France. Received July 27, 1998; revision requested October 14; revision received January 6, 1999; accepted March 26. Address reprint requests to Y.C. (e-mail: ctouzery@dijon.fnclcc.fr).
 |
Abstract |
|---|
 TOP Abstract IntroductionMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
MATERIALS AND METHODS: Sixty-one patients were examined after a myocardial infarction. The whole left ventricle was imaged by using 5-mm contiguous breath-hold cine MR short-axis sections with no gap (SAng) (two-dimensional fast low-angle shot sequence, 9/4.8 [repetition time msec/echo time msec]). The effect of omitting one section in two (short-axis sections with 5-mm gap [SA5mm]) or two sections in three (short-axis sections with 10-mm gap [SA10mm]) was studied.
RESULTS: In the comparison of SA5mm or SA10mm with respect to the reference SAng, the standard error of the estimate (SEE) for the diastolic volume did not exceed the 6.1% interobserver SEE, and the SEE for the ejection fraction remained lower than the 3% interobserver SEE. The measured wall thickness was not affected. In addition, six simple geometric models were compared with SAng and yielded an SEE of 9.5%–28.1% for the diastolic volume and 3.8%–13.3% for the ejection fraction.
CONCLUSION: In the study of left ventricles with heterogeneous contractility, short-axis imaging is more accurate than geometric modeling and permits wall thickness measurements when an intersection gap of 5 or 10 mm is used.
Index terms: Heart, ejection fraction, 51.91 • Heart, MR, 51.121412, 51.12142 • Heart, ventricles, 524.91 • Heart, volume, 51.92 • Magnetic resonance (MR), cine study, 51.12142 • Magnetic resonance (MR), volume measurement, 51.121412, 51.12142 • Myocardium, infarction, 511.771
|
Introduction |
|---|
|
TOPAbstractIntroductionMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
Magnetic resonance (MR) imaging can acquire images in oblique imaging planes, for example, in the short and long axes of the left ventricle. The accuracy of cine MR imaging for assessing left ventricular function has been proved in the results of several studies. After a myocardial infarction, MR imaging appears to be an excellent method for evaluating ejection fraction, myocardial wall thickness and thickening, and regional contractile dysfunction in the left ventricle (1). Many studies (2–5) show the superior accuracy and precision of MR imaging. Moreover, MR imaging coupled with dobutamine administration may become a choice technique for patients who have ischemic heart disease. The acquisition of a data set covering the entire left ventricle at several levels of dobutamine infusion requires a rapid but highly reproducible technique. In most of the current studies in which dobutamine perfusion is used, only three to four imaging planes were used (6–10).The aim of this work was to determine the extent to which the three-dimensional data set could be reduced without substantially modifying the calculated left ventricular volumes and the systolic and diastolic wall thicknesses, especially in ventricles with heterogeneous contractile properties.
|
MATERIALS AND METHODS |
|---|
|
TOPAbstractIntroductionMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
The study was conducted in accordance with the recommendations of the local ethics committee, and oral informed consent was obtained from each patient and from each healthy subject.
Imaging Techniques
MR imaging was performed by using a breath-hold technique, a 1.5-T whole-body MR imager (Magnetom Vision; Siemens, Erlangen, Germany), and a phased-array body coil. The breath-hold cine MR imaging data were acquired by using an electrocardiographically gated gradient-echo sequence: segmented two-dimensional fast low-angle shot with view sharing, 9/4.8 (repetition time msec/echo time msec), flip angle of 15°, nine lines per segment, a field of view of 300–400 mm, and a section thickness of 5 mm. The acquisition matrix varied from 108 x 256, acquired in 12 cardiac cycles, to 144 x 256 lines, acquired in 16 cardiac cycles, depending on the ability of patients to maintain breath hold.
The patients were allowed to choose the most comfortable breath-hold position (ie, inhalation or expiration). Each patient usually repeated the same kind of breath hold during the entire study.
The use of view-sharing techniques improved the temporal resolution to 50 msec per frame (11). Hence, a set of multiphase images was acquired for a single imaging plane during a single breath hold of less than 15 seconds.
Phantom Study
The design of the phantom was deliberately simplistic, and the aim of the phantom study was to provide precise and accurate distance and surface measurements from images generated by an imaging sequence that incorporated the view-sharing technique. It was therefore necessary to verify that the measured values were truly representative of the real spatial dimensions present in the phantom. Consequently, a phantom formed by two concentric cylinders of Plexiglas was constructed. The phantom was filled with a solution of 0.1 mmol/L of gadolinium(III) chloride, or Gd3+, which corresponds to a T1 value of approximately 1.5 seconds—the approximate T1 of blood in vivo at 1.5 T. The cylinder wall thickness was 1 mm in each case, which left a 21-mm space between the cylinders. For convenience, the phantom was positioned such that the cylinder axes were perpendicular to the principal axis of the MR imager. Images were therefore acquired in the coronal plane. The same two-dimensional fast low-angle shot sequence was used for patient studies and for measurements of the phantom (Fig 1).
|
Data Analysis
Two independent observers (Y.C., F.G.) determined the outer and inner contours of the phantom on a single section. Each observer repeated the tracings 30 times. The imaging parameters described for human studies were applied for the calculation of the thickness and the inner volume of the phantom on the selected section.
For the patients and healthy subjects, the endocardial and epicardial borders of the end-diastolic and end-systolic images were manually traced on every section. For endocardial contour tracing, the papillary muscles, if present, were ignored. One of the two observers (Y.C.) determined the contours of all the patients and healthy subjects, and the data from this observer were used in the final analysis. For the purpose of reproducibility assessment, the second observer (F.G.) participated in the contouring of 20 randomly selected patient data sets and 10 healthy subject data sets.
Left ventricular end-diastolic volume, end-systolic volume, and ejection fraction were calculated from short-axis views. The volumes were calculated as the sum of the cavity areas multiplied by the section interval (section thickness + section gap). Three different intersection gaps were studied: short-axis sections with no gap (SAng), short-axis sections with 5-mm gap (SA5mm), and short-axis sections with 10-mm gap (SA10mm).
When appropriate, the missing volume at the base of the ventricle was assumed to be equivalent to a cylinder of an area equal to the surface of the most basal section and a height equal to the distance between this section and the diastolic or systolic valve plane. The missing apical volume was approximated by a semiellipsoid of a short-axis area equal to the surface of the most apical section and a long-axis radius equal to the distance between the most apical section and the cavity extremity.
Volumes also were calculated by using several geometric methods: the modified Simpson rule model, the hemisphere-cylinder model, the biplane ellipsoid model, the horizontal long-axis plane and vertical long-axis plane models, and the Teicholz model (Fig 2) (12).
|
The sections were further grouped according to three levels: basal, median, distal. For the SAng set, the average thickness was calculated from all the acquired sections. For the SA5mm set, the average was calculated after retaining every other section starting from the most basal section. For the SA10mm set, the same method was applied after retaining only one of every three sections. Because in most cases the number of sections retained was not a multiple of three (ie, the number of levels), a weighted average method of thickness calculation was used to take into account without bias the data from sections that overlapped two levels.
The ventricular apex accounted for the 13th segment. The wall thickness measurements relative to the apex were calculated from the average of the horizontal long-axis plane and the vertical long-axis plane data. For the purposes of matching myocardial segments with coronary distribution from angiography, the anterior wall, septum, and apex were assigned to the left anterior descending arterial territory, the lateral wall was assigned to the left circumflex coronary artery, and the inferoposterior wall was assigned to the right coronary artery.
Radionuclide Ventriculography
All patients underwent radionuclide ventriculography within 7 days after the MR imaging examination, and the left ventricular ejection fraction was calculated by a clinician (F.B.) blinded to the MR imaging results.
The study was performed with patients in the supine position by using blood pool labeled with 740 MBq (20 mCi) of technetium 99m. Data were acquired in a left anterior oblique view. Images with 6 million counts were obtained by using 32 frames and a 64 x 64 matrix with a pixel size of 2.4 mm. Cardiac cycles with R-R intervals not within 10% of the average were rejected. The resting left ventricular ejection fraction was determined by means of manual definition of end-diastolic, end-systolic, and background regions.
Coronary Angiography
All patients underwent coronary angiography. The severity of stenosis was determined by one observer (J.E.W.) who used the software (Digital Cardiac Imaging; Philips Medical, Eindhoven, the Netherlands) as part of the routine catheterization protocol. Severe coronary disease was defined as stenosis of at least 70% of the diameter in one major epicardial artery. On ventriculograms obtained in the two perpendicular planes (60° left anterior oblique and 30° right anterior oblique views), the segments with normal kinetics were classified as normal, whereas the hypokinetic and akinetic segments were classified as abnormal.
Statistical Analysis
For the statistical analysis, the results obtained from the contiguous short-axis data (SAng) were considered the standard of reference. For the left ventricular volumes and ejection fractions, the comparison of the reduced short-axis data sets or the geometric model results with the reference method was performed by means of linear regression analysis. The standard error of the estimate (SEE), the coefficient of correlation (r), and the associated P value were reported. The same analysis was applied to the comparison of the reduced short-axis data sets or geometric model results with the radionuclide ventriculographic results.
The evaluation of the reduced short-axis data sets or geometric models for ejection fraction determination was performed by using the SEE of regression between these methods and the reference. These methods were ranked according to increasing SEE.
Only methods giving an SEE less than the SEE of reproducibility were considered valid. The comparison of the SEEs was performed by using an F test on the variances.
The agreement between methods for assessment of ejection fraction, end-diastolic or end-systolic thicknesses, or absolute thickening has been illustrated with Bland-Altman plots (13). The comparison of the wall thickness results between segments in healthy subjects, patient normal segments, and patient abnormal segments was performed by using a Student t test. Statistical significance was accepted for P values less than or equal to .05. Parameter reproducibility between two observers was assessed by means of a linear regression analysis.
|
RESULTS |
|---|
|
TOPAbstractIntroductionMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
Patient Data
Coronary angiograms.—Coronary angiography demonstrated three-vessel disease in 10 patients, two-vessel disease in 21 patients, and one-vessel disease in 30 patients.
Volumes and ejection fraction analysis.—The results of the regression analysis (r value, SEE, and P value) for the interobserver comparison of volumes and ejection fraction from the SAng data are listed in Table 1.
|
|
|
|
Thickness parameter analysis.—The mean thicknesses in 20 patients and 10 healthy subjects was calculated by using SAng, and the results of the regression analysis of the two observer measurements (r value and SEE) are listed in Table 4. The dispersion of differences of end-diastolic wall thicknesses, end-systolic wall thicknesses, or absolute wall thickening calculated according to either SA5mm or SA10mm and the contiguous-section method SAng are illustrated by the Bland-Altman plots in Figure 4.
|
|
|
|
|
|
DISCUSSION |
|---|
|
TOPAbstractIntroductionMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
The use of contiguous 5-mm-thick sections renders MR imaging one of the most precise methods for imaging the whole left ventricle. The choice of thin sections has not been made at the expense of the imaging quality because the phased-array coil provides a much higher signal-to-noise ratio than the standard body coil. The implementation of the view-sharing technique in the sequence provided an improved temporal resolution of 50 msec, instead of 80 msec for the conventional sequence (11,15).
The main finding of the present study is that an intersection gap of 5–10 mm does not induce a significant loss of accuracy in the measurements of both volumes and ejection fractions. The methods of calculation, including one or two long-axis planes, exhibit lower correlations. The SEE remains lower than the interobserver variability of the reference method (3%) for only the short-axis methods with an intersection gap of 5 and 10 mm. Rather than considering the 3% ejection fraction interobserver variability, the interstudy variability might have been used to define the limits of acceptable variability. Many studies (3–5,18–20) in which echocardiography, radionuclide angiography, and MR imaging were used have shown a 3%–7% interstudy variability owing to various physiologic contributions such as ventricular afterload, hydration state, or heart rate. If such a large tolerance range were to be used, only the modified Simpson rule with a SEE of only 3.8% would appear to be an acceptable technique among the six geometric models.
The comparison of ejection fraction calculated by using MR imaging with ejection fraction calculated by using equilibrium-gated angioscintigraphy leads to a similar conclusion concerning the respective value of the geometric models. The lower values of ejection fraction recorded at radionuclide angiography probably reflect the slight underestimation of this parameter with use of radionuclide techniques (21).
In patients with hypertrophic cardiomyopathy with homogeneous contraction of the left ventricle, Dulce et al (12) found that the modified Simpson rule and biplane ellipsoid models showed the highest correlations with the analysis of a three-dimensional data set. Biplane angiocardiography has been validated for measurement of left ventricular volumes (22). Single-plane ventriculography demonstrated good accuracy for measurement of ventricular volume and ejection fraction (23). These long-axis approaches have been used in the determination of the left ventricular function by using MR imaging with either modest correlations (16,24) or a good agreement (25–27) with contrast-enhanced ventriculography. The mean left ventricular ejection fraction of 34% reported in the patients after infarction in the present study indicates a severe contraction heterogeneity, which might explain the discrepancy between the long-axis views and the short-axis measurements. These discrepancies may be because the cine-ventriculographic image of the left ventricle is a projection of the whole cavity and not of an isolated tomographic plane (28).
Myocardial thickness measurements of all the myocardial segments is only possible from the short-axis approach. The small intermethod thickness difference induced in short-axis methods by including an intersection gap remained within the interobserver variation of the reference SAng method. Although lower than the 9.0–10.4 mm findings of Baer et al (29) for the diastolic wall thickness in healthy volunteers, our results show no significant difference according to the territory of the coronary arteries. These results are in the 6–10 mm thickness range found by Sechtem et al (30) in a work based on spin-echo 10-mm-thick transverse sections and are consistent with the thicknesses of 7.40–7.66 mm obtained by Holman et al (31) using 10-mm-thick sections and a three-dimensional approach. As shown in previous studies, the absence of a major thinning in the infarcted area (31,32) could be due to the fact that our patients were examined within 15 days after the onset of the myocardial infarction, whereas Baer et al (29,33) examined their patients at least 4 months after myocardial infarction.
By using myocardial tagging methods, the through-plane motion has been shown to have an average value ranging from 12.8 mm at the base to 1.6 mm at the apex (34,35). Even if tagging sequences had been added to compensate for the misregistration, other distortion phenomena would not have been corrected easily, such as warping and tilting of the short-axis plane during heart contraction (35,36). In view of all these complex movements of the heart, averaging the chords on a limited number of 13 segments as proposed in the present study should contribute to minimizing the misregistration errors.
Breath holding also might have induced some variations in the ventricular volumes, as suggested by echocardiographic findings (37). However, Sakuma et al (15) showed the lack of a significant difference in the left ventricular volumes in MR imaging acquisitions obtained with or without breath holding. In the method described in this study, breath-hold acquisition is required to avoid important image artifacts induced by respiratory motion when using the phased-array coil.
The described approach was limited by examinations with unsatisfactory quality in nine of 70 patients. This is mainly because of poor quality cardiac gating owing to electrocardiographic artifacts in the magnet or by claustrophobia. Only one patient could not sustain the required breath holds.
To limit the duration of breath hold, two approaches may be considered: The first is to reduce the number of phase-encoding lines with the risk of diminishing the spatial resolution of the images, and the second is to increase the number of lines per phase segment with an obvious decrease in the temporal resolution. In the present study, we chose to optimize imaging with respect to temporal resolution (38,39).
Another limitation of our study is the labor-intensive manual processing. Many semiautomatic methods have been proposed (40,41) for the calculation of wall thickness and ventricular volumes. Unfortunately, no commercially available package allows reliable measurements, especially for wall thickness, without considerable user intervention.
There have been some recent attempts to use black-blood imaging with an improved wall definition (42). However, for the time being, these techniques do not allow comprehensive coverage of either space (the entire heart) or time (the complete cardiac cycle).
The results of the present study show that left ventricular imaging based on short-axis imaging of the heart from the base to the apex with use of 5-mm sections separated by an intersection gap of 5–10 mm provides an accurate measurement of the left ventricular volume, ejection fraction, and myocardial thickness. This method meets with the requirements for studying the effects of stress on the left ventricle by using MR imaging. Foreseeable improvements in the speed of acquisition, in gating, and in postprocessing should make this approach an attractive alternative to other techniques for the noninvasive study of left ventricular function.
|
Acknowledgments |
|---|
|
Footnotes |
|---|
Author contributions: Guarantor of integrity of entire study, F.B.; study concepts, Y.C.; study design, F.B.; definition of intellectual content, P.L.; literature research, Y.C., O.R.; clinical studies, Y.C., J.E.W., A.L., S.R., P.M.W.; experimental studies, P.M.W.; data acquisition, Y.C., F.G., A.L.; data analysis, Y.C., C.T., P.M.W.; statistical analysis, C.T.; manuscript preparation, Y.C., C.T.; manuscript editing, P.M.W.; manuscript review, J.E.W.
|
References |
|---|
|
TOPAbstractIntroductionMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
This article has been cited by other articles:
|
|
 |
|
  M. Tumkosit, C. G. Martin, E. Bayram, T. M. Morgan, K. S. Lane, P. Rerkpattanapipat, C. A. Hamilton, K. M. Link, and W. G. Hundley Left Ventricular Spherical Remodeling and Apical Myocardial Relaxation: Cardiovascular MR Imaging Measurement of Myocardial Segments Radiology, August 1, 2007; 244(2): 411 - 418. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. Fenchel, A. M. Scheule, N. I. Stauder, U. Kramer, K. Tomaschko, T. Nagele, C. Bretschneider, H.-P. Schlemmer, C. D. Claussen, and S. Miller Atherosclerotic Disease: Whole-Body Cardiovascular Imaging with MR System with 32 Receiver Channels and Total-Body Surface Coil Technology--Initial Clinical Results Radiology, December 1, 2005; 238(1): 280 - 291. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
K. U. Juergens, M. Grude, D. Maintz, E. M. Fallenberg, T. Wichter, W. Heindel, and R. Fischbach Multi-Detector Row CT of Left Ventricular Function with Dedicated Analysis Software versus MR Imaging: Initial Experience Radiology, February 1, 2004; 230(2): 403 - 410. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 K. U. Juergens, M. Grude, E. M. Fallenberg, C. Opitz, T. Wichter, W. Heindel, and R. Fischbach Using ECG-Gated Multidetector CT to Evaluate Global Left Ventricular Myocardial Function in Patients with Coronary Artery Disease Am. J. Roentgenol., December 1, 2002; 179(6): 1545 - 1550. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
S. Miller, O. P. Simonetti, J. Carr, U. Kramer, and J. P. Finn MR Imaging of the Heart with Cine True Fast Imaging with Steady-State Precession: Influence of Spatial and Temporal Resolutions on Left Ventricular Functional Parameters Radiology, April 1, 2002; 223(1): 263 - 269. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
Ai x I. b, Left ventricular volume determined by using the modified Simpson rule model. Am = area of the left ventricular cavity measured on the short-axis section at the mitral valve level, Ap = area of the left ventricular cavity measured on the short-axis section at the level of the papillary muscle, L = length of the left ventricular long axis. Left ventricular volume (V) was calculated according to the following equation: V = [Am + (Am + Ap)/2 + Ap/3] x L/3. c, Left ventricular volume determined by using the hemisphere-cylinder model. Am = area of the left ventricular cavity measured on the short-axis section at the mitral valve level, L = length of the left ventricular long axis. Left ventricular volume (V) was calculated according to the following equation: V = (Am + 2 x Am/3) x L/2. d, Left ventricular volume determined by using the biplane ellipsoid model. A1 = area of the left ventricular cavity measured on the vertical long-axis section, A2 = area of the left ventricular cavity measured on the horizontal long-axis section, L = length of the left ventricular long axis. Left ventricular volume (V) was calculated according to the following equation: V = 8 x A1 x A2/(3
L). e, Left ventricular volume determined by using either the horizontal long-axis plane ellipsoid model or the vertical long-axis plane ellipsoid model. A1 = area of the left ventricular cavity measured on the vertical long-axis section or the horizontal long-axis section, L = length of the left ventricular long axis. Left ventricular volume (V) was calculated according to the following equation: V = 8 x A12/(3
