| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Cardiac Imaging |
1 From Medical Clinic I (T.P., T.S.) and Dept of Radiology (M.S., C.K.B.), Univ Hosp of Mannheim, Theodor-Kutzer-Ufer 1-3, 68167 Mannheim, Germany; and Dept of Cardiology, VU Univ Medical Ctr, Amsterdam, the Netherlands (H.P.K., O.B., A.B., M.M.B.H., A.C.v.R.). Received Apr 8, 2004; revision requested Jun 18; revision received Jul 30; accepted Aug 25. T.P. supported by a grant from the Medical Faculty of the Univ of Mannheim. H.P.K. supported in part by grants from the Medical Faculty of the Rheinisch-Wesfälische Technische Hochschule, Aachen, Germany, and the Grimmke-Stiftung, Düsseldorf, Germany. O.B. supported by grant 2001.158 from the Netherlands Heart Foundation. Address correspondence to T.P. (e-mail: theano.papavassilliu{at}med.ma.uni-heidelberg.de).
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
MATERIALS AND METHODS: The study was approved by the local ethics committee, and each subject gave informed consent before participating. Twenty healthy subjects and 20 consecutive patients underwent 1.5-T cardiovascular MR imaging. Seven to 12 short-axis views encompassing the entire LV were acquired by using true FISP. Endocardial and epicardial contours were traced manually. The data sets in each patient were analyzed twice: with inclusion of endocardial trabeculae in the LV cavity volume and with exclusion of endocardial trabeculae from the cavity volume. On the basis of these two contour sets, the end-diastolic (ED) and end-systolic (ES) LV volumes, ejection fraction (EF), and LV mass were calculated. Additionally, interobserver and interexamination reproducibility was assessed by using Bland-Altman analysis.
RESULTS: Compared with exclusion of trabeculae, inclusion of trabeculae in the LV cavity volume resulted in significantly larger ED and ES LV volumes (mean differences, 21 mL ± 11 [standard deviation] and 19 mL ± 33, respectively; P < .001) and lower EFs (mean difference, –2% ± 2; P < .001). The calculated LV mass was significantly smaller with inclusion than with exclusion of trabeculae (mean difference, –21 g ± 12; P < .001). All interobserver and interexamination limits of agreement based on inclusion of trabeculae, except those for EF measurements, were superior to those based on exclusion of trabeculae. At measurement reproducibility comparisons, differences in interobserver ED LV volume and LV mass and interexamination LV mass were statistically significant, favoring the inclusion of trabeculae in the LV cavity volume.
CONCLUSION: Trabeculae significantly affect quantifications of LV volume and mass. The superior reproducibility of LV measurements with the inclusion of endocardial trabeculae in the cavity volume favors this tracing algorithm for clinical use.
© RSNA, 2005
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
Cardiovascular magnetic resonance (MR) imaging performed by using conventional cine gradient-echo (ie, breath-hold segmented fast low-angle shot) sequences is an established reference standard for the assessment of LV volume and mass (4–6). Recent technical improvements have enabled the implementation of an alternative to the gradient-echo cardiovascular MR imaging pulse sequence: true fast imaging with steady-state precession (FISP) (7). The true FISP sequence has become the technique of choice for assessing regional and global cardiac function because it yields substantially improved blood-tissue contrast and high temporal and spatial resolution (8,9). Use of this technique facilitates excellent endocardial border definition throughout the entire cardiac cycle and easier delineation of the borders, particularly in areas affected by slow blood flow, such as the regions around the papillary muscles and the LV trabeculae (10). Moreover, true FISP MR imaging allows visualization of even small endocardial trabeculae, which often could not be depicted by using conventional gradient-echo sequences owing to the lower blood-myocardium contrast and the lower spatial resolution.
The summation-of-sections method is the established reference standard for assessing LV volume and mass. At present, however, there is no clear consensus as to which tracing method for outlining the papillary muscles and the endocardial trabeculae should be used to achieve the most reliable and reproducible results. Thus, the aims of our study were to prospectively assess the effect of including or excluding endocardial trabeculae in LV measurements and to evaluate the reproducibility of these measurements at cine cardiovascular MR imaging performed with true FISP.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
Image Acquisition
All MR imaging examinations were performed by using a 1.5-T whole-body system (Magnetom Sonata; Siemens Medical Systems, Erlangen, Germany). A dedicated four-element, phased-array cardiac coil was used. The MR images were acquired during repeated end-expiratory breath holds. Scout MR images were obtained to plan the acquisition of the final double-oblique long- and short-axis views. Electrocardiographically gated cine MR images were then acquired by using a segmented steady-state precession sequence—true FISP—and the following parameters: 3.2/1.2 (repetition time msec/echo time msec), 35-msec temporal resolution, 1.4 x 1.8-mm in-plane spatial resolution, and 5-mm section thickness. Seven to 12 short-axis views that were 1 cm apart and encompassed the entire LV were acquired. The time required to obtain the short-axis sections ranged between 10 and 15 minutes.
Image Analysis and Determination of LV Parameters
The MR images were transferred to a separate workstation (Sun Sparcstation; Sun Microsystems, Mountain View, Calif). Image analysis was performed by using the MASS software package (Medis, Leiden, the Netherlands), as previously reported (11). The cine loops were reviewed, and the end-diastolic (ED) and end-systolic (ES) frames were identified for each short-axis section position. The ED frame was defined as the frame showing the largest cavity area, and the ES frame was defined as the frame showing the smallest cavity area. The most basal section was the section that at ED and ES still showed a wall thickness that was compatible with the LV myocardium and that extended over at least 50% of the myocardial circumference. At ES, this most basal section could also show a part of the LV outflow tract or the mitral valve leaflets. The most basal section could differ by one section position between ED and ES.
In each patient, the first observer (T.P., with 4 years experience in cardiac MR imaging) determined the ED and ES frames, as well as the number and end-section position of short-axis sections used for the analysis, according to the described criteria. This information was subsequently used by the second observer (M.S., with 3 years experience in cardiac MR imaging).
Epicardial and endocardial contours were outlined manually on each ED and ES short-axis-view MR image. For each patient and image section, the contrast and brightness settings were optimized to achieve the best possible contrast between the myocardium and the LV cavity. The papillary muscles were outlined separately and included in the myocardial mass. The papillary muscles were identified on the long- and short-axis MR images, and their location and extent were correlated with each other. On the basal sections, the papillary muscles were usually disconnected from the LV wall and could be easily identified. On the more apical sections, where the papillary muscles were attached to the wall, it was sometimes difficult to clearly differentiate these muscles from the trabeculae. When the papillary muscles could not be clearly distinguished from the trabeculae, they were treated as trabeculae.
Right ventricular trabeculations arising from the interventricular septum and the epicardial fat were excluded from the LV mass. So that the effect of trabeculae on LV parameter measurements could be specifically assessed and the variability associated with manual tracing of the epicardial borders and the papillary muscles could be minimized, we changed neither the boundaries of the epicardium nor the contours of the papillary muscles during the tracing of the endocardium with either method (with trabeculae included and with trabeculae excluded). However, to minimize observer bias, there was a time interval of at least 1 week between the two tracing convention–based analyses of each data set. Thus, only the contours of the epicardial boundaries and of the papillary muscles were stored. The following tracing conventions were used:
Inclusion of trabeculae in LV cavity volume.—Trabeculum was defined as a structure 1.5 mm or larger in diameter that was protruding into the LV cavity and was attached to the LV wall. During tracing of the endocardial borders, in a separate window there was always a continuous movie display of the section being evaluated so that the trabeculae could be visually followed during contraction and relaxation. This display enabled better differentiation of the LV trabeculae from the LV wall, particularly on the ES frames. During endocardial tracing, trabeculae were included in the LV cavity volume. This inclusion resulted in a smooth endocardial contour that was almost parallel to the epicardial contour. An example of tracing performed by using this inclusion (outer-contour) convention is depicted in Figure 1a.
|
|
On the basis of these two contour sets, the ED and ES LV volumes, EF, and LV mass at ED were calculated. ED and ES LV volumes were calculated by using a modification of the Simpson rule. The EF was calculated as follows: EF = [(EDV – ESV)/EDV] · 100%, where EDV is the ED LV volume and ESV is the ES LV volume. The LV mass (MLV) was calculated from the diastolic phase as follows: MLV = 1.05 · (Vepi – Vendo), where Vepi is the epicardial volume and Vendo is the endocardial volume. The average image analysis time was 40 minutes per subject (ie, per group of subject images).
To assess interobserver measurement reproducibility, a second independent and blinded observer (M.S.) repeated the LV measurements in each data set by using the same two tracing conventions. This observer was blinded to the subject details (ie, patient names, ages, and medical histories) and to the first observer's findings. For assessment of interexamination measurement variability, six subjects returned for repeat MR imaging at least 1 week after the first examination.
Statistical Analyses
Data are presented as means ± SDs. The two tracing algorithms were compared in terms of LV volume and mass measurements by using the paired t test. Agreement between the two tracing algorithms was assessed by using the method described by Bland and Altman (12). Interobserver measurement reproducibility also was assessed by using Bland-Altman analysis, which yielded the mean difference, the limits of agreement of the mean difference (difference ± 1.96 SDs), and the corresponding 95% confidence intervals. Moreover, the coefficients of repeatability for each tracing algorithm and for the preceding parameters (ED volume, ES volume, EF, and LV mass) were assessed, with the coefficient of repeatability calculated as two times the SDs of the differences between the two measurements. Similarly, interexamination measurement reproducibility was assessed by using Bland-Altman analysis. In addition, the coefficient of variability (equal to the SD of the difference between the two measurements divided by the mean of the two measurements, expressed as a percentage) was calculated.
The statistical significance of differences in interobserver and interexamination reproducibility was assessed by using an extension of the Bland-Altman method. A log transformation of the squared differences between the two measurements was performed. If the squared difference was zero, it was replaced by half of the next smallest value. A two-tailed paired t test of the logged squared differences for the two tracing algorithms (12,13) could then be performed. At all comparisons, P < .05 was considered to indicate a significant difference.
Power calculations to assess the sample sizes needed in this study could not be performed in advance because no estimate of the SD of the difference between the outer- and inner-contour tracing algorithms was available. However, after acquiring data from our first 10 subjects, we did perform a power calculation (two-sided paired t test performed with SAS, version 8.2 software [SAS Institute, Cary, NC]) on the basis of the SD estimated from these data. According to this calculation, the number of subjects needed to establish a power of 90% (with 
= .05) was 12 for the comparison between the outer- and inner-contour tracing conventions performed by one observer. The sample sizes needed to establish a power of 90% for the assessment of interobserver agreement on measurements obtained by using outer-contour tracing (ie, ED and ES LV volumes, LV mass, and EF) and a power of 80% for the assessment of interobserver agreement on measurements obtained by using inner-contour tracing were 13 and 40 subjects at maximum, respectively.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
|
|
|
|
|
|
|
|
|
|
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
The study results demonstrate that in patients and healthy subjects, LV trabeculae significantly affect all LV measurements. Compared with the inclusion of trabeculae in the LV cavity volume, the exclusion of endocardial trabeculae resulted in systematically smaller LV volumes and larger EF and LV mass values.
There currently is no clear consensus as to which method of tracing the papillary muscles and endocardial trabeculae should be used to achieve the most reliable and reproducible results. Several investigators have reported normal cardiovascular MR imaging indexes of LV volume and mass that were obtained by using different tracing conventions. To our knowledge, Lorenz et al (16) reported the first normal ranges for LV mass and volume that were obtained by using a conventional free-breathing gradient-echo sequence. Because of signal intensity averaging, a clear delineation of the endocardial trabeculae was not possible with use of this technique. In another study involving the use of a gradient-echo sequence with breath holding, Marcus et al (11) reported normal LV volume and mass values. There was a difference in the ED LV mass values obtained in the Lorenz et al and Marcus et al studies, however. This difference may have been partially related to differences in data acquisition technique (breath holding vs no breath holding) and study subject age range between the two studies. A third main difference between the two studies was related to the drawn contour of the endocardial border: Marcus et al (11) drew a larger contour around the endocardial border, which resulted in lower LV mass values compared with those in the Lorenz et al (16) study.
A study by Moon et al (13), in which FISP and fast low-angle shot MR image acquisitions were compared for the assessment of LV volume, mass, and function, revealed small but significant differences in ED and ES LV volumes and LV mass between the two techniques. There were two reasons for these differences: The endocardial contour was drawn larger and the epicardial contour was drawn smaller on the FISP MR images. On fast low-angle shot MR images, the papillary muscles can appear larger and confluent with the myocardium. Moreover, blood between the trabeculations depicted on fast low-angle shot images can appear to be a part of the myocardium and thereby result in lower LV volumes and higher LV mass values compared with the values calculated by using FISP images. These differences in measurements between conventional gradient-echo and steady-state precession pulse sequences were also confirmed in the recently published study of Alfakih et al (17).
The importance of including or excluding the endocardial trabeculae and the papillary muscles in LV measurements obtained with conventional gradient-echo pulse sequences was previously emphasized by Ibrahim et al (18) and Marcus et al (19). Ibrahim et al used cine breath-hold gradient-echo cardiovascular MR imaging to measure the LV volumes, EF, and LV mass in 52 healthy subjects and found that the inclusion of the papillary muscles in the blood pool, as compared with the exclusion of the papillary muscles and the trabeculae, resulted in systematically higher ED and ES LV volumes and lower LV mass values (P < .001). The exclusion of trabeculae from the blood pool resulted in the highest EF values (P < .001). The findings in our study are consistent with these results.
Marcus et al (19) measured the LV mass in 40 healthy subjects and quantified the portion of the mass that was contributed by the most basal section and by the endocardial trabeculae. The results indicated that the mean contribution to the ED LV mass from the papillary muscles was 6 g ± 5, or 4% ± 3. It was most striking that the ES LV mass was larger than the ED LV mass by a mean of 5 g ± 9, or 3% ± 6, when the most basal section and the papillary muscles were included in the LV mass measurements. Marcus et al hypothesized that the endocardial trabeculae were responsible for the apparent increase in mass during systole.
The endocardial border is not a sharply demarcated line between blood and myocardium; rather, it is a trabecular zone with a complex border. This factor leads to clinically relevant differences between measurements obtained by using different tracing conventions, as described earlier. Thus, comparing published LV measurement results is very difficult. Additionally, published normal ranges of cardiovascular MR imaging–based LV mass and volume values cannot be easily adopted because different tracing conventions were used. These differences emphasize the need for uniform criteria for outlining the endocardial contour. Functional true FISP cardiac MR imaging, as compared with conventional gradient-echo MR imaging, yields a high level of endocardial detail and allows easier discrimination of the endocardial borders. Therefore, true FISP cardiovascular MR imaging may become the technique of choice for assessing LV volume and mass.
The clinical value of a specific analysis method is determined on the basis of not only the accuracy but also the reproducibility of the method. Therefore, we sought to determine whether one of the two described tracing conventions was superior to the other in terms of measurement reproducibility. The interobserver and interexamination reproducibility of LV volume and mass measurements obtained by including endocardial trabeculae in the LV cavity volume was superior to that of measurements obtained by excluding endocardial trabeculae. There are two main reasons for these differences: (a) Because of their small size, trabeculae are difficult to differentiate from the LV wall, and, thus, tracing is complicated. (b) The decision of which endocardial trabeculae to exclude from the cavity volume is very challenging and varies considerably, even when experienced observers perform the tracing.
On the basis of our observations, we suggest that endocardial trabeculae be excluded from the myocardial mass and included in the LV cavity volume. This tracing convention may be easier to follow for different observers and thus result in less observer variability. Thus, the superior interobserver and interexamination reproducibility of measurements obtained by including endocardial trabeculae in the LV cavity volume favors this tracing algorithm for clinical use.
Another advantage of including trabeculae in the LV cavity volume is that measurements of regional wall thickness and systolic wall thickening can be expected to be more realistic. Peters et al (20) found that measurements of myocardial wall strain derived from tagged MR images showed correlation between regions of the trabeculae and the papillary muscles and regions of high strain, and this correlation led to an overestimation of the function in the lateral wall.
There were limitations to our study. First, during systole, the trabeculae and the papillary muscles are less well defined owing to compression and folding. In the current study, the compressed trabeculae appeared as a segment of continuous myocardial tissue in some regions at ES, and this made accurate delineation of the endocardial borders very challenging. Therefore, a movie display of the section being evaluated was always available. This display allowed us to differentiate the trabeculae from the LV wall at ES by integrating the visual information obtained continuously during the cardiac cycle. Nevertheless, identifying the trabecular tissue at ES remained difficult in some cases, which might support the use of the convention with which endocardial trabeculae are excluded from the LV cavity volume. In these cases, the inner-contour convention allowed easier delineation of the endocardial border. This finding explains the similar coefficients of reproducibility for measurements of the ES LV volume and the EF.
Second, the best tracing algorithm is probably that which is the most accurate and the most reproducible. Because we measured LV parameters in vivo, we were not able to test for accuracy. Therefore, we sought to assess whether one of the two described tracing conventions was superior to the other in terms of reproducibility. Third, the sample size of examined subjects in our study was relatively small; however, the sample represented a population with a wide spectrum of cardiac disorders.
In conclusion, endocardial trabeculae significantly affect the quantification of LV volume and mass. The superior reproducibility of LV measurements obtained by including endocardial trabeculae in the cavity volume favors this tracing algorithm for clinical use. Our findings may aid in reaching a consensus on uniform tracing criteria to standardize cardiovascular MR imaging measurements of LV parameters.
|
FOOTNOTES |
|---|
Abbreviations: ED = end diastole • EF = ejection fraction • ES = end systole • FISP = fast imaging with steady-state precession • LV = left ventricle • SD = standard deviation
Authors stated no financial relationship to disclose.
Author contributions: Guarantors of integrity of entire study, T.P., A.C.v.R.; study concepts, T.P., H.P.K., A.C.v.R., M.S.; study design, T.P., H.P.K., A.C.v.R., T.S., A.B., C.K.B.; literature research, T.P., O.B., T.S., C.K.B.; clinical studies, T.P., A.C.v.R.; data acquisition, T.P., H.P.K., A.C.v.R., M.M.B.H., O.B., A.B., M.S., C.K.B.; data analysis/interpretation, T.P., H.P.K., T.S.; statistical analysis, T.P., M.M.B.H., C.K.B., T.S.; manuscript preparation, T.P., H.P.K., A.C.v.R., M.S.; manuscript definition of intellectual content, A.B., M.M.B.H., T.S., C.K.B., O.B.; manuscript editing and final version approval, T.P., H.P.K., A.C.v.R.; manuscript revision/review, all authors
|
References |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
This article has been cited by other articles:
|
|
 |
|
  N. C. F. Codella, J. W. Weinsaft, M. D. Cham, M. Janik, M. R. Prince, and Y. Wang Left Ventricle: Automated Segmentation by Using Myocardial Effusion Threshold Reduction and Intravoxel Computation at MR Imaging Radiology, September 1, 2008; 248(3): 1004 - 1012. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
S. Uribe, T. Tangchaoren, V. Parish, I. Wolf, R. Razavi, G. Greil, and T. Schaeffter Volumetric Cardiac Quantification by Using 3D Dual-Phase Whole-Heart MR Imaging Radiology, August 1, 2008; 248(2): 606 - 614. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 V. Mor-Avi, C. Jenkins, H. P. Kuhl, H.-J. Nesser, T. Marwick, A. Franke, C. Ebner, B. H. Freed, R. Steringer-Mascherbauer, H. Pollard, et al. Real-Time 3-Dimensional Echocardiographic Quantification of Left Ventricular Volumes: Multicenter Study for Validation With Magnetic Resonance Imaging and Investigation of Sources of Error J. Am. Coll. Cardiol. Img., July 1, 2008; 1(4): 413 - 423. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. Kankaanpaa, H.-R. Lehto, J. P. Parkka, M. Komu, A. Viljanen, E. Ferrannini, J. Knuuti, P. Nuutila, R. Parkkola, and P. Iozzo Myocardial Triglyceride Content and Epicardial Fat Mass in Human Obesity: Relationship to Left Ventricular Function and Serum Free Fatty Acid Levels J. Clin. Endocrinol. Metab., November 1, 2006; 91(11): 4689 - 4695. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
