| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Technical Developments |
1 From the British Heart Foundation Cardiac MRI Unit (D.R.M., G.J.B., K.A., T.R.J., S.P., M.U.S.) and Department of Medical Physics (J.P.R.), Leeds General Infirmary, Leeds, England. Received February 9, 2004; revision requested April 20; revision received May 10; accepted June 15. D.R.M. supported by a Marie Curie research grant from the European Commission. Address correspondence to D.R.M., Cardiac MRI Team, Franz-Volhard-Klinik, Humboldt Universität, Charité Campus Buch, Wiltbergstrasse 50, 13125 Berlin, Germany (e-mail: messroghli@fvk-berlin.de).
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMaterials and Methods ResultsDiscussionREFERENCES |
|---|
© RSNA, 2005
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMaterials and Methods ResultsDiscussionREFERENCES |
|---|
In contrast, regional LV parameters are usually evaluated from a limited number of short-axis sections by using standardized segmentation models. These sections can be acquired either after the performance of a multisection short-axis study or selectively. In situations where regional rather than global parameters are the focus of interest, a comprehensive data set covering the entire LV is unnecessary, and the time spent on the acquisition and handling of superfluous data could be saved.
Selective data acquisition for the evaluation of regional LV parameters at rest and at stress has been performed in a number of MR imaging studies in which three, four, or five short-axis sections were used, with section positions derived from either systolic or diastolic long-axis images (3–8). To our knowledge, none of these approaches have been systematically validated.
The consensus statement published by the Cardiac Imaging Committee of the Council on Clinical Cardiology of the American Heart Association has provided guidelines for nomenclature and localization of the standard segments of the LV (9). It recommends the use of a 17-segment model and provides anatomic landmarks to be considered for the selection of sections for display.
The purpose of this study was to evaluate the accuracy of a simple selective short-axis section-positioning approach compared with the established criteria for cardiac section positioning and with the conventional approach using multisection coverage of the entire LV.
|
Materials and Methods |
|---|
|
TOPABSTRACTINTRODUCTIONMaterials and Methods ResultsDiscussionREFERENCES |
|---|
Twenty-one subjects (11 men and 10 women; mean age, 32 years ± 11 [standard deviation]) were healthy volunteers with no history of cardiovascular disease, a normal resting blood pressure, and a normal 12-lead electrocardiogram.
Twenty-three subjects (21 men and two women; mean age, 56 years ± 12) were patients with a history of acute (n = 21) or chronic (n = 2) myocardial infarction as diagnosed according to clinical, enzymatic, and electrocardiogram changes. Patients were recruited consecutively from March 2003 until September 2003. Patients with arrhythmia, with general exclusion criteria for MR imaging (eg, pacemaker, other non–MR-compatible metal implants, claustrophobia), or who were clinically unstable were not enrolled. In two additional patients, MR imaging had to be aborted because of claustrophobia before any cine images could be acquired, and both patients were excluded from the study. An electrocardiogram helped define the infarct location as anterior or anteroseptal in 13 cases, as inferior in six, and as posterolateral in four. Table 1 shows subject characteristics in more detail.
|
Positioning was performed by using the three-dimensional graphical-positioning interface of the MR imager in the following manner: After the acquisition of transverse localizer images, a two-chamber cine image was derived by positioning a parasagittal section through the LV apex and the middle of the mitral annulus. On the basis of the two-chamber view, a basal short-axis view was planned parallel to the mitral annulus. From the resulting short-axis image, a four-chamber cine image (the section intersecting the midpoint of the LV and the junction of the free and the inferior walls of the right ventricle) was obtained (10).
For the selective short-axis method, end systole was identified in the two-chamber and four-chamber view cine studies by selecting the image with the smallest discernable LV cavity. On these systolic images, a stack of five parallel sections was positioned in the short-axis orientation between the mitral annulus and the apex of the LV (Fig 1, step A). The intersection gaps were increased in 0.5-mm increments, and the stack was recentered to closely fit the LV so that the outer margin of the most apical section reached the outer boundary of the LV apex and the proximal border of the most basal section aligned with the mitral annulus (Fig 1, step B; Fig 2). Of these five section positions, only the middle three sections were used for actual imaging (repetition time msec/echo time msec, 3.48/1.74; flip angle, 55°; section thickness, 8 mm; pixel size, 1.8 x 2.3 mm; 18 phases; heart phase interval, 30–67 msec), which led to data acquisition in basal, midcavity, and apical section positions (Fig 1, step C).
|
|
At the completion of the aforementioned protocol (and after being removed from the imaging table), all volunteers underwent a repeat imaging protocol with new plan images and a repeat selective short-axis study to determine the interstudy reproducibility of the method. This second MR imaging session was performed by a different technologist (G.J.B. or T.R.J.) blinded to the results of the first study.
Assessment of Accuracy of Section Positioning
To determine the accuracy of section positioning achieved in normal hearts, all of the volunteer selective short-axis cine images were evaluated visually, section by section, and independently by two observers (D.R.M. and K.A.), who had 5 and 4 years of experience with cardiac MR imaging, respectively. The criteria used were derived from the American Heart Association recommendations for the positioning of short-axis sections published by Cerqueira et al (9). The precise instructions for use in echocardiography and the general principles related to tomographic imaging were adopted as guidelines. The full standard criteria used are given in Table 2. In the event of a disagreement between the observers, a consensus judgment was made following joint review of the images. Compliance for each criterion was calculated as the percentage fraction of cases where the images matched the criterion. The multisection short-axis images were analyzed similarly, and compliance was noted for the three sections that most closely matched the criteria for basal, midcavity, and apical locations. Percentage agreement between the results of the two repeated selective short-axis studies and between the results of selective short-axis and multisection short-axis studies was calculated. 
Statistics for interstudy and selective short-axis—multisection short-axis agreement were calculated from the complete data sets.
|
1. The intersection gap was chosen by the two independent technologists to fit the stack of five sections to the systolic long-axis length of the LV. Since the section thickness was always 8 mm, the intersection gap should be directly proportional to the systolic long-axis length (Fig 1).
2. The end-diastolic LV diameter in the midcavity short-axis section was measured in two different orientations corresponding to the vertical long axis and horizontal long axis. To determine the starting points for the measurements, the insertion of the anterior free wall of the right ventricle to the LV was taken as a reference point, and six radial myocardial segments starting from there were defined (Fig 3). Any variability in orientation between the two data sets should result in a different angulation of the short-axis sections relative to the heart axis, leading to an angulated cross section of the LV. This in turn would be reflected in the LV diameters.
|
To determine the accuracy of section positioning achieved with the selective short-axis method in hearts with abnormal anatomy and function, both selective short-axis and multisection short-axis studies from all patients were evaluated qualitatively by the same two observers using the same methods as described for the volunteers.
After the completion of all qualitative analyses, all studies with rejected criteria were reassessed by both of the observers to determine the causes for rejection.
To provide an estimate of the degree of global LV dysfunction in our patient group, end-diastolic volumes and ejection fractions were derived for both study populations from the multisection short-axis studies. Endocardial and epicardial contours of end-diastolic and end-systolic images from the full sets of sections were traced where myocardium was visible in at least 50% of the circumference of the LV, and a modified Simpson rule (13) was used for the calculations. All evaluations were performed by using commercially available software (MASS 5.0; Medis, Leiden, the Netherlands).
Statistical Analysis
For all quantitative data, Pearson correlation analysis and Student t tests were performed by using SPSS for Windows (release 11.5.0; SPSS, Chicago, Ill). Independent-sample t tests were used for comparing volunteers and patients, and paired t tests were used for comparing results from initial and repeat studies. P < .05 was used to indicate significant differences. Except for end-diastolic volumes and ejection fractions, Bland-Altman analysis (14) was used to assess agreement between different quantitative measurements by using software (Analyze-it 2002; Analyze-it Software, Leeds, UK). The same software program was used for the calculation of scores to estimate agreement in the qualitative assessments.
|
Results |
|---|
|
TOPABSTRACTINTRODUCTIONMaterials and Methods ResultsDiscussionREFERENCES |
|---|
Accuracy of Section Positioning in Normal Hearts
Figure 4 shows a representative set of images acquired in a volunteer by using the selective short-axis method. Table 3 summarizes the two-observer consensus results for the accuracy of section positioning as compared with the criteria for standard sections by using the selective short-axis and the multisection short-axis approach. Both approaches showed a similar good overall compliance with the visual criteria (90.9% and 92.1%, respectively). Compliance was lowest with both approaches for criterion 2 (78.6% and 76.2%, respectively).
|
|
= 0.76. Overall agreement between selective short axis and multisection short axis was 94.5%, with = 0.67. In terms of individual criteria, interstudy agreement was lowest for criterion 1 (85.7%). This was also the criterion with the lowest rate of interobserver agreement regarding the judgment of criteria compliance (88.1%; overall, 95.2%; data not shown in Table 3). The quantitative evaluation of interstudy reproducibility showed very good agreement between the two selective short-axis studies performed by the two different technologists. Results regarding the agreement of intersection gap chosen for the two repeated studies and of midcavity end-diastolic LV diameters in vertical long-axis and horizontal long-axis projections are summarized in Table 4. Figure 5 depicts the correlation of intersection gaps and end-diastolic LV diameters between the two repeated studies.
|
|
|
= 0.55), but criterion 2 was frequently not met when either technique was used (compliance in 47.8% and 60.9%, respectively). The assessments of the selective short-axis sections by the two observers agreed in the majority of cases (overall, 94.2%; 87.0% for criterion 2; data not shown in Table 6). The results of both techniques for criterion 2 agreed in 69.6% of cases (overall, 91.3%). Retrospective analysis of the causes for rejection of criterion 2 showed that the tips of the papillary muscles were visible to some extent in all but one of the basal sections.
|
|
|
Discussion |
|---|
|
TOPABSTRACTINTRODUCTIONMaterials and Methods ResultsDiscussionREFERENCES |
|---|
As opposed to the conventional multisection short-axis approach, the intersection gap with the selective short-axis method is variable and is dependent on the systolic LV long-axis length. Positioning with the selective short-axis method is therefore performed on relative levels rather than at absolute distances within the LV. This ensures that the same segments of the LV are selected for imaging in a follow-up study even if there were substantial changes of LV morphology, such as dilatation due to remodeling after myocardial infarction (5).
The criteria for short-axis sections applied in this study were derived from a consensus statement recently published by the American Heart Association Writing Group on Myocardial Segmentation and Registration for Cardiac Imaging (9). Since the statements regarding short-axis section orientation for MR imaging appeared to us to be not sufficiently detailed for practical section positioning in routine imaging, we also adopted for our study the more precise recommendations for echocardiography. We further clarified the requirements for the three short-axis sections before the beginning of the study to provide clear guidelines for qualitative assessment (Table 2). Overall compliance of the selective short-axis method with the positioning criteria was good in both volunteers and patients and was comparable to the values achieved with the multisection short-axis approach. Use of criterion 1 (visibility of the tips of the mitral valve leaflets) showed the lowest rate of agreement between the two observers since true valve appearances had to be differentiated from flow artifacts. This may be due to the limited ability of MR imaging in depicting thin fast-moving structures at present because of insufficient temporal resolution.
Compared with multisection short-axis data sets with eight to 14 sections, acquisition of only the three required sections leads to a relative reduction of image acquisition time by 63%–79%, which is independent of the type of cine pulse sequence being used. At the same time, the selective short-axis method satisfies the principle recommendation of the American Heart Association guidelines: "For regional analysis of left ventricular function or myocardial perfusion, the left ventricle should be divided into equal thirds perpendicular to the long axis of the heart" (9). Basal, midcavity, and apical sections are always evenly spaced with the selective short-axis method, which prevents an overrepresentation of any particular section.
Positioning in end systole ensured that the basal section contained the LV throughout the whole cardiac cycle, which resulted in agreement with criterion 3 ("no LV outflow tract visible in basal section") in 95.2% of volunteers and 100% of patients. If imaging had been based on diastolic positioning, the basal section may have included parts of the LV outflow tract during systole in subjects with good systolic ventricular function. In contrast, our approach may theoretically lead to a position of the basal section that may be marginally too basal in patients with severely reduced systolic function. This could be seen as a reason why the rate of compliance with criterion 2 (tips of papillary muscles fully visible at end diastole in basal section) was particularly low in the patient group. Indeed, since compliance with criterion 2 was low with both the selective short-axis and the multisection short-axis approach, it is more likely that the strict rules defined by this criterion cannot always be met by tomographic means, especially when LV anatomy and function are impaired.
The high level of interstudy agreement of the quantitative parameters representing the section orientation can be explained by the simplicity of the positioning scheme. It is based on the identification of only two anatomic landmarks (mitral valve level and apex of the LV), which minimizes the possible sources of error.
The good results shown for interstudy, interobserver, and intraobserver variability in end-diastolic wall thickness, end-systolic wall thickness, and wall motion of the midcavity section underline the robustness of the selective short-axis approach for the quantification of regional wall motion parameters. Apart from quantitative measurements, selective short-axis positioning should favorably complement semiquantitative approaches such as the American Society of Echocardiography score (15), as recently modified for cardiac MR imaging by Sierra-Galan et al (16).
Since the selective short-axis approach is based on position planning in systolic images, accuracy of positioning is optimized for systolic images. Nevertheless, significant differences on diastolic images of follow-up studies are only to be expected when systolic function (not LV volume) has changed substantially between studies. Otherwise, the degree of through-plane motion during the cardiac cycle, and therefore the displacement of the imaging section, should remain unaltered because of the dynamic variation of the intersection gap in our method. For the same reason, the selective short-axis positioning approach should also be applicable to other forms of cardiac MR imaging studies that require standard short-axis localization, such as first-pass perfusion imaging, as long as the temporal positioning during the cardiac cycle is taken into consideration and is consistent.
This study had limitations. The exclusion of segment 17 (the tip of the apex) is an inherent drawback of the selective short-axis method, since it uses the short-axis orientation. Segment 17 should be ideally evaluated from long-axis views (9), where it does not move substantially through the imaging plane or has partial volume effects.
Our study was designed to validate the accuracy of selective short-axis section positioning in normal hearts and in hearts with myocardial infarction, which was performed in a qualitative fashion. To assess quantitatively the reproducibility of the results, repeat studies were performed in volunteers but not in patients, with the assumption that reproducibility, other than accuracy, is influenced only by factors that were the same in both groups (positioning of the subject within the imager bore, section positioning technique, contour drawing) and not by abnormal anatomy.
As with any selective section positioning technique, it is theoretically possible that very small myocardial defects might be missed with the selective short-axis method if they are located in the gap between two sections. However, it is beyond the scope of this study to determine the minimum size of functional defects to be detected with the selective short-axis method.
In summary, imaging of the central three of five selective short-axis sections positioned between mitral valve annulus and apex in systolic long-axis views leads to accurate and reproducible basal, midcavity, and apical short-axis sections representing the standard LV segments one through 16. The selective short-axis method can therefore serve as a robust approach for the assessment of regional LV function in cardiac MR imaging. The selective short-axis approach is easy to integrate into existing imaging protocols and requires only minimal technical input and expertise.
|
ACKNOWLEDGMENTS |
|---|
|
FOOTNOTES |
|---|
Authors stated no financial relationship to disclose.
Author contributions: Guarantor of integrity of entire study, D.R.M.; study concepts, D.R.M., K.A., S.P., J.P.R., M.U.S.; study design, D.R.M., T.R.J., S.P., J.P.R., M.U.S.; literature research, D.R.M.; clinical studies, D.R.M., G.J.B., T.R.J.; data acquisition, D.R.M., G.J.B., T.R.J.; data analysis/interpretation, D.R.M., K.A.; statistical analysis, D.R.M., K.A.; manuscript preparation, definition of intellectual content, and editing, D.R.M.; manuscript revision/review and final version approval, all authors
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMaterials and Methods ResultsDiscussionREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  P. G. Masci, S. Dymarkowski, F. E. Rademakers, and J. Bogaert Determination of Regional Ejection Fraction in Patients with Myocardial Infarction by Using Merged Late Gadolinium Enhancement and Cine MR: Feasibility Study Radiology, January 1, 2009; 250(1): 50 - 60. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
D. R. Messroghli, S. Plein, D. M. Higgins, K. Walters, T. R. Jones, J. P. Ridgway, and M. U. Sivananthan Human Myocardium: Single-Breath-hold MR T1 Mapping with High Spatial Resolution--Reproducibility Study Radiology, March 1, 2006; 238(3): 1004 - 1012. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
