AJR AJR-based Continuing Ed for Technologists
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     

This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF)
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Right arrow Citation Map
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Juergens, K. U.
Right arrow Articles by Fischbach, R.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Juergens, K. U.
Right arrow Articles by Fischbach, R.
Social Bookmarking
Add to CiteULike   Add to Complore   Add to Connotea   Add to Del.icio.us   Add to Digg   Add to Reddit   Add to Technorati  
What's this?
Hotlight (NEW!)
Right arrow
What's Hotlight?
DOI:10.2214/AJR.07.2283
AJR 2008; 190:308-314
© American Roentgen Ray Society

Original Research

Automated Threshold-Based 3D Segmentation Versus Short-Axis Planimetry for Assessment of Global Left Ventricular Function with Dual-Source MDCT

Kai Uwe Juergens1, Harald Seifarth1, Felix Range2, Susanne Wienbeck1, Mirja Wenker1, Walter Heindel1 and Roman Fischbach1,3

1 Department of Clinical Radiology, University of Muenster, Albert-Schweitzer-Straße 33, D-48149 Muenster, Germany.
2 Department of Cardiology and Angiology, University of Muenster, Muenster, Germany.
3 Department of Radiology, Asklepios Clinic Altona, Hamburg, Germany.

Received March 20, 2007; accepted after revision August 1, 2007.

 
Address correspondence to K. U. Juergens (kujuerg{at}uni-muenster.de).


Abstract
Top
 Abstract
Introduction
 Subjects and Methods
 Results
 Discussion
 References
 
OBJECTIVE. The purpose of this study was to evaluate software for threshold-based 3D segmentation of the left ventricle in comparison with traditional 2D short axis–based planimetry (Simpson method) for measurement of left ventricular (LV) volume and global function with state-of-the-art dual-source CT.

SUBJECTS AND METHODS. Fifty patients with known or suspected coronary artery disease underwent coronary CT angiography. LV end-diastolic, end-systolic, and stroke volumes and ejection fraction were determined from axial images to which 3D segmentation had been applied and from short-axis reformations from 2D planimetry. Interobserver variability was assessed for both approaches.

RESULTS. Threshold-based 3D LV segmentation had excellent correlation with 2D short-axis results (end-diastolic volume, R = 0.99; end-systolic volume, R = 0.99; stroke volume, R = 0.90; ejection fraction, R = 0.97; p < 0.0001). Bland-Altman analyses revealed systematic underestimation of LV end-diastolic volume (–7.4 ± 8.9 mL) and LV end-systolic volume (–7.0 ± 4.4 mL) with the 3D segmentation approach and 2.8 ± 3.3% overestimation of LV ejection fraction. Interobserver variation with 3D segmentation analysis was significantly (p < 0.001) less (e.g., LV ejection fraction, 0.1 ± 1.7%) than with the 2D technique, and mean analysis time was significantly shorter (172 ± 20 vs 248 ± 29 seconds; p < 0.05).

CONCLUSION. Automated threshold-based 3D segmentation enables accurate and reproducible dual-source CT assessment of LV volume and function with excellent correlation with results of 2D short-axis analysis. Exclusion of papillary muscles from LV volume results in small systematic differences in quantitative values.

Keywords: dual-source CT • left ventricular function • short axis–based planimetry • Simpson method • threshold 3D segmentation algorithm


Introduction
Top
 Abstract
 Introduction
Subjects and Methods
 Results
 Discussion
 References
 
Left ventricular (LV) volume and myocardial mass are independent predictors of morbidity and mortality among patients with coronary artery disease, and global LV function is considered the strongest determinant of heart failure and death due to myocardial infarction. Accurate and reproducible determination of LV volumetric and functional parameters is important for clinical diagnosis and risk stratification in the care of patients with suspected or documented heart disease [1–3]. ECG-gated MDCT of the heart provides retrospective quantitative information on LV volume changes throughout the cardiac cycle and consecutive information on global LV function. Global LV functional parameters measured with MDCT have been found to be in good agreement with results of cine MRI and transthoracic echocardiography [4–6]. MDCT studies of LV function have been performed with 2D planimetry of short-axis CT image reformations, the so-called Simpson method [7–12], as introduced with cine MRI studies for LV functional analysis. This 2D approach has the limitation that numerous user interactions are time consuming and introduce systematic error in definition of the short-axis plane and determination of the endocardial contours and most basal and apical sections [13, 14].

Threshold-based 3D volumetry with a region-growing segmentation approach has been introduced as an alternative to traditional short axis–based planimetry for LV volume measurement with MDCT. Clinical experience with the 3D approach has been limited to two studies [15, 16] of preliminary software tools that were conducted with small numbers of patients. We therefore evaluated the use of commercially available 3D segmentation software in comparison with the traditional 2D Simpson method with regard to correlation and systematic differences and interobserver variability in state-of-the-art dual-source CT of a cohort of patients with known or suspected coronary artery disease.


Subjects and Methods
Top
 Abstract
 Introduction
 Subjects and Methods
Results
 Discussion
 References
 
Patients
Fifty consecutively registered patients (16 women, 34 men; mean age, 57.4 ± 12.1 years; age range, 25–85 years) with known or suspected coronary artery disease were referred for CT coronary angiography for noninvasive evaluation of the coronary arteries, determination of coronary artery plaque burden, or assessment of coronary artery bypass graft patency. Exclusion criteria for CT were known allergy to iodine-containing contrast media, renal insufficiency (calculated glomerular filtration rate ≤ 60 mL/min/1.73 m2), thyroid disorder, pregnancy, and hemodynamic instability. The study was approved by the institutional review board, and the patients gave written informed consent for the CT protocol.

CT Protocol and Image Acquisition
All CT examinations were performed with a dual-source CT system (Somatom Definition, Siemens Medical Solutions) according to the routine protocol for coronary CT angiography. Oral and IV β-adrenoreceptor antagonists were not administered before CT irrespective of heart rate. CT data were acquired in a craniocaudal direction. The scanner settings were detector configuration, 32 x 0.6 mm; slice acquisition, 64 x 0.6 mm with a z-flying focal spot for each detector; gantry rotation time, 330 milliseconds; constant temporal resolution, 83 milliseconds; pitch, 0.21–0.43 automatically adapted to the heart rate; tube voltage, 120 kV for each tube; tube current–time product, 340 mAs/rotation. The ECG pulsing window was individually set to 60–80% of R-R interval for patients with a heart rate of 51–70 beats/min and 30–80% of R-R interval for heart rates exceeding 70 beats/min. ECG data were digitally recorded during acquisition and were stored with the CT data set.

The scan range covered the entire heart from the level of the tracheal bifurcation to the diaphragm. A 70-mL bolus of nonionic contrast material (iomeprol 350 mg I/mL) followed by 40 mL of contrast material diluted to 40% concentration with 0.9% saline solution and a 50-mL saline chaser was continuously injected into an antecubital vein through an 18-gauge catheter at a constant injection rate of 5 mL/s (Stellant D CT injector, Medrad). The scan delay was controlled with bolus tracking whereby a region of interest was positioned centrally in the ascending aorta, and 150 H was used as the target for triggering the CT scan.

CT Protocol and Image Acquisition
Dual-source CT image reconstruction for axial and short-axis reformations was performed on the scanner workstation (Syngo CT-2007 A, VA 10A, Siemens Medical Solutions) in 5% steps, dividing the R-R interval into 20 phases for each patient. Axial images had a 2-mm section thickness and a 1-mm reconstruction increment. Short-axis images had an 8-mm section thickness and zero gap and covered the heart from the mitral valve to the apex, the standard approach used in several studies [7–12, 14]. The field of view was set to 180 mm2 for all image series.

CT Data Analysis
Axial images were transferred to an off-line workstation (Aquarius, TeraRecon). Assessment of LV function was performed with a dedicated software package for cardiac function analysis (TVA version 3.5.2.1, TeraRecon) with threshold-based region-growing 3D segmentation of the LV cavity. After all phases of the cardiac cycle were loaded, the software generated long-axis and short-axis displays of the heart. The mitral valve plane had to be manually defined in the horizontal and vertical long-axis planes. The contrast-filled LV lumen was then automatically segmented for all 20 cardiac phases. The lower segmentation threshold was individually adapted in instances in which automatic threshold definition did not result in homogeneous depiction of the contrast-enhanced LV lumen. The software was used to calculate LV end-diastolic volume (LVEDV); LV end-systolic volume (LVESV), LV stroke volume (LVSV), or LVEDV – LVESV; and LV ejection fraction (LVEF), or LVSV/LVEDV x 100.

Short-axis CT image reformations were evaluated on another off-line workstation (Leonardo, Siemens Medical Solutions) with a standard software package for cardiac CT (Syngo Argus, Siemens Medical Solutions). The end-diastolic and end-systolic phases were visually determined as the phases with the largest and smallest, respectively, ventricular cavities. Endocardial borders were contoured semi-automatically and checked visually for accuracy on all CT images with a discernible LV cavity. The most basal slice was defined as the image closest to the mitral valve annulus showing LV myocardium in at least 50% of its perimeter. The most apical image was the last image with a visible LV lumen. Papillary muscles were included as part of the LV cavity. The software provided LVEDV, LVESV, LVSV, and LVEF.

LV functional analysis with both approaches was performed by a reader with 8 years of experience in cardiac radiology (reader 1). For assessment of interobserver variability, a first-year resident in general radiology (reader 2), who had been trained with 10 cases on each workstation, measured results for 23 patients using each method. To blind both readers to the results of the first evaluation, a time delay of 4 weeks after analysis of each patient's axial data sets was observed before short-axis measurement. The time required to load the reconstructed CT data sets into the software interface and to complete the 2D and 3D data analyses was recorded for readers 1 and 2.

Statistical Analysis
Continuous variables were expressed as mean ± SD, median, and minimum and maximum values. The nonparametric Mann-Whitney U test for independent samples was used to test for significant differences in LV volumes and functional parameters with the axial and short-axis approaches. A value of p < 0.05 was considered statistically significant. Pearson's correlation coefficient and Bland-Altman analyses [17] were performed to determine linear correlation and to calculate limits of agreement and systematic errors for each pair of values of LVEDV, LVESV, LVSV, and LVEF between ventricular segmentation and planimetry as measured by reader 1. Interobserver variability of LVEDV, LVESV, LVSV, and LVEF was assessed by calculation of absolute differences and performance of Bland-Altman analysis. All computations were performed with commercially available software (MedCalc 9.2.0.1, MedCalc Software).


Results
Top
 Abstract
 Introduction
 Subjects and Methods
 Results
Discussion
 References
 
All 50 patients completed dual-source CT coronary angiography without complications. The mean heart rate during the CT studies was 69.3 ± 13.9 beats/min (median, 68 beats/min; range, 50–110 beats/min). Three-dimensional segmentation of the LV cavity and short axis–based planimetry were feasible for all patients.

Mean analysis time with the 2D short-axis approach was 248 ± 29 seconds (reader 1). Threshold-based 3D segmentation analysis was performed within the significantly shorter mean time of 172 ± 20 seconds (p < 0.05). Likewise, the analysis time required by reader 2 was significantly higher for 2D short-axis analysis (293 ± 34 seconds) in comparison with the 3D segmentation algorithm (190 ± 22 seconds; p < 0.05). Comparison of the readers revealed a significantly longer duration of 2D short-axis analysis (p > 0.05) by reader 2 versus reader 1. Differences in 3D analysis time were not significantly different for the two readers. With use of the protocol described for contrast administration, the mean attenuation value in the LV cavity was 292.4 ± 50.6 H (range, 189–464 H). The mean attenuation value in the myocardial septum was 84.2 ± 12.3 H (p < 0.001).

The results of LV volume and global functional measurements are summarized in Tables 1, 2, 3. LV volumes and LVEF measured with 3D segmentation by reader 1 had excellent correlation with the results with 2D short-axis planimetry (Fig. 1A, 1B, 1C, 1D) for all variables analyzed (RLVEDV = 0.99, RLVESV = 0.99, RLVSV = 0.90, RLVEF = 0.97; p < 0.0001). Bland-Altman analysis revealed systematic underestimation of LVEDV (–7.4 ± 8.9 mL) and LVESV (–7.0 ± 4.4 mL) with use of 3D segmentation and slight overestimation (2.8 ± 3.3%) of LVEF (Fig. 2A, 2B, 2C, 2D).


View this table:
[in this window]
[in a new window]

  		TABLE 1: Left Ventricular Volume and Global Functional Measurements in Dual-Source CT of Patients (n = 50) with Known or Suspected Coronary Artery Disease: 3D Segmentation Algorithm (Reader 1)

 

View this table:
[in this window]
[in a new window]

  		TABLE 2: Left Ventricular Volume and Global Functional Measurements in Dual-Source CT of Patients (n = 50) with Known or Suspected Coronary Artery Disease: Short-Axis-Based Planimetry (Reader 1)

 

View this table:
[in this window]
[in a new window]

  		TABLE 3: Left Ventricular Volume and Global Functional Measurements in Dual-Source CT of Patients (n = 50) with Known or Suspected Coronary Artery Disease: 3D Segmentation Algorithm Versus Short Axis–Based Planimetry (Reader 1)

 

Figure 1
View larger version (8K):
[in this window]
[in a new window]
[as a PowerPoint slide]
  		Fig. 1A —Results of linear regression analysis 3D threshold segmentation versus short-axis planimetry (reader 1). Scatter diagram shows results for left ventricular end-diastolic volume (y = 2.1766, x = 0.9666).

 

Figure 2
View larger version (8K):
[in this window]
[in a new window]
[as a PowerPoint slide]
  		Fig. 1B —Results of linear regression analysis 3D threshold segmentation versus short-axis planimetry (reader 1). Scatter diagram shows results for left ventricular end-systolic volumes (y = 4.3000, x = 0.9582).

 

Figure 3
View larger version (9K):
[in this window]
[in a new window]
[as a PowerPoint slide]
  		Fig. 1C —Results of linear regression analysis 3D threshold segmentation versus short-axis planimetry (reader 1). Scatter diagram shows results for left ventricular ejection fraction (y = 2.1539, x = 1.0103).

 

Figure 4
View larger version (9K):
[in this window]
[in a new window]
[as a PowerPoint slide]
  		Fig. 1D —Results of linear regression analysis 3D threshold segmentation versus short-axis planimetry (reader 1). Scatter diagram shows results for left ventricular stroke volume (y = 7.6274, x = 0.9086).

 

Figure 5
View larger version (13K):
[in this window]
[in a new window]
[as a PowerPoint slide]
  		Fig. 2A —Results of Bland-Altman analysis depict systematic error and limits of agreement for 3D threshold segmentation and short axis–based planimetry (reader 1). Plot shows results for left ventricular end-diastolic volume.

 

Figure 6
View larger version (13K):
[in this window]
[in a new window]
[as a PowerPoint slide]
  		Fig. 2B —Results of Bland-Altman analysis depict systematic error and limits of agreement for 3D threshold segmentation and short axis–based planimetry (reader 1). Plot shows results for left ventricular end-systolic volume.

 

Figure 7
View larger version (13K):
[in this window]
[in a new window]
[as a PowerPoint slide]
  		Fig. 2C —Results of Bland-Altman analysis depict systematic error and limits of agreement for 3D threshold segmentation and short axis–based planimetry (reader 1). Plot shows results for left ventricular stroke volume.

 

Figure 8
View larger version (13K):
[in this window]
[in a new window]
[as a PowerPoint slide]
  		Fig. 2D —Results of Bland-Altman analysis depict systematic error and limits of agreement for 3D threshold segmentation and short axis–based planimetry (reader 1). Plot shows results for left ventricular ejection fraction.

 

Interobserver agreement was excellent for threshold-based 3D segmentation. The mean absolute differences between reader 1 and 2 were 9.9 ± 5.9 mL (LVEDV), 3.7 ± 1.9 mL (LVESV), 6.6 ± 4.7 mL (LVSV), and 1.4 ± 0.9% (LVEF). Bland-Altman analysis resulted in systematic differences of 9.1 ± 7.1 mL (LVEDV), 3.1 ± 2.7 mL (LVESV), and –0.1 ± 1.7% (LVEF) (Fig. 3A, 3B, 3C). For 2D short-axis planimetry, the mean absolute differences between readers were 14.7 ± 8.8 mL (LVEDV), 10.0. ± 9.5 mL (LVESV), 12.9 ± 9.7 mL (LVSV), and 6.9 ± 5.7% (LVEF). Bland-Altman analysis revealed systematic differences of 7.9 ± 15.1 mL (LVEDV), –3.0 ± 13.7 mL (LVESV), and 4.3 ± 7.6% (LVEF). All differences determined with 2D short-axis analysis were significantly higher (p < 0.001) than those with the 3D segmentation approach.


Figure 9
View larger version (12K):
[in this window]
[in a new window]
[as a PowerPoint slide]
  		Fig. 3A —Results of Bland-Altman analysis depict systematic error and limits of agreement between reader 1 and reader 2 in use of 3D threshold segmentation. Plot shows results for left ventricular end-diastolic volume.

 

Figure 10
View larger version (11K):
[in this window]
[in a new window]
[as a PowerPoint slide]
  		Fig. 3B —Results of Bland-Altman analysis depict systematic error and limits of agreement between reader 1 and reader 2 in use of 3D threshold segmentation. Plot shows results for left ventricular end-systolic volume.

 

Figure 11
View larger version (12K):
[in this window]
[in a new window]
[as a PowerPoint slide]
  		Fig. 3C —Results of Bland-Altman analysis depict systematic error and limits of agreement between reader 1 and reader 2 in use of 3D threshold segmentation. Plot shows results for left ventricular ejection fraction.

 

Discussion
Top
 Abstract
 Introduction
 Subjects and Methods
 Results
 Discussion
References
 
Global LV function is a powerful independent predictor of morbidity and mortality irrespective of the cause of myocardial dysfunction, and LV volumes and myocardial mass are independent predictors among patients with coronary heart disease. Accurate and reproducible determination of LV myocardial function is essential for clinical diagnosis, risk stratification, and treatment planning in the care of patients with suspected or documented coronary heart disease [1–3].

Because cardiac MDCT studies are performed with retrospective ECG gating and data acquisition throughout the cardiac cycle, imaging information on any phase of the cardiac cycle can be obtained without additional radiation exposure, allowing evaluation of LV functional and anatomic characteristics. Global LV functional measurements obtained with 4- to 64-MDCT and short axis–based planimetry (Simpson method) are evidence of good agreement with findings on cine MRI, echocardiography, catheter ventriculography, and gated SPECT [7–16, 18–28]. High attenuation differences between myocardium and contrast agent in combination with near isotropic CT data allow threshold-based segmentation of LV volume. LV cavity segmentation can be performed with little user interaction, allowing fast estimation of global LV functional parameters. Several companies have developed software tools for LV functional analysis with 3D segmentation. Clinical experience, however, is scarce, and validation against the traditional approach of LV planimetry is restricted, to our knowledge, to only two studies with small patient numbers [15, 16].

Threshold-based 3D segmentation algorithms require homogeneous opacification of the LV and right ventricular cavities and a large difference in the attenuation values of the LV lumen and myocardial wall to enable reliable delineation of the endocardial and epicardial contours, including the intraventricular septum. If the duration of injection of the highly concentrated bolus is too long, the resulting hyperenhancement and inhomogeneous opacification of the right ventricular cavity can impair automatic contour detection and LV and right ventricular segmentation. Therefore, correct timing of both contrast phases of the CT coronary angiography protocol is crucial to optimize contrast bolus planning. A preliminary study [16] of analysis software in which a similar segmentation approach was used showed that proper LV segmentation was feasible in only two thirds of the patients. In our study with 50 patients with an LVEF between 24.7% and 78.1% (Fig. 4A, 4B, 4C, 4D, 4E, 4F), LV segmentation was feasible in all cases. In the LV cavity, the mean attenuation value of 292 ± 50 H revealed a significant difference of 208 H compared with mean attenuation values measured in the intraventricular myocardial septum. This finding is in concordance with the results of the previous study [16], in which a mean difference of 205 H was reported for the group of patients with proper segmentation of the left ventricle. By contrast, the patients with insufficient LV segmentation had a significantly lower mean difference of only 120 H [16].


Figure 12
View larger version (106K):
[in this window]
[in a new window]
[as a PowerPoint slide]
  		Fig. 4A —68-year-old man with three-vessel coronary artery disease and inferolateral myocardial infarction. Diastolic (A–C) and systolic (D–F) CT image reformations show reduced diastolic wall thickness and absence of systolic wall thickening of inferior and inferolateral walls of left ventricle. Left ventricular volumes and ejection fraction determined with 3D threshold segmentation algorithm (A, B, D, E) and 2D short axis–based planimetry (Simpson method) (C, F) show comparable results.

 

Figure 13
View larger version (79K):
[in this window]
[in a new window]
[as a PowerPoint slide]
  		Fig. 4B —68-year-old man with three-vessel coronary artery disease and inferolateral myocardial infarction. Diastolic (A–C) and systolic (D–F) CT image reformations show reduced diastolic wall thickness and absence of systolic wall thickening of inferior and inferolateral walls of left ventricle. Left ventricular volumes and ejection fraction determined with 3D threshold segmentation algorithm (A, B, D, E) and 2D short axis–based planimetry (Simpson method) (C, F) show comparable results.

 

Figure 14
View larger version (76K):
[in this window]
[in a new window]
[as a PowerPoint slide]
  		Fig. 4C —68-year-old man with three-vessel coronary artery disease and inferolateral myocardial infarction. Diastolic (A–C) and systolic (D–F) CT image reformations show reduced diastolic wall thickness and absence of systolic wall thickening of inferior and inferolateral walls of left ventricle. Left ventricular volumes and ejection fraction determined with 3D threshold segmentation algorithm (A, B, D, E) and 2D short axis–based planimetry (Simpson method) (C, F) show comparable results.

 

Figure 15
View larger version (103K):
[in this window]
[in a new window]
[as a PowerPoint slide]
  		Fig. 4D —68-year-old man with three-vessel coronary artery disease and inferolateral myocardial infarction. Diastolic (A–C) and systolic (D–F) CT image reformations show reduced diastolic wall thickness and absence of systolic wall thickening of inferior and inferolateral walls of left ventricle. Left ventricular volumes and ejection fraction determined with 3D threshold segmentation algorithm (A, B, D, E) and 2D short axis–based planimetry (Simpson method) (C, F) show comparable results.

 

Figure 16
View larger version (74K):
[in this window]
[in a new window]
[as a PowerPoint slide]
  		Fig. 4E —68-year-old man with three-vessel coronary artery disease and inferolateral myocardial infarction. Diastolic (A–C) and systolic (D–F) CT image reformations show reduced diastolic wall thickness and absence of systolic wall thickening of inferior and inferolateral walls of left ventricle. Left ventricular volumes and ejection fraction determined with 3D threshold segmentation algorithm (A, B, D, E) and 2D short axis–based planimetry (Simpson method) (C, F) show comparable results.

 

Figure 17
View larger version (71K):
[in this window]
[in a new window]
[as a PowerPoint slide]
  		Fig. 4F —68-year-old man with three-vessel coronary artery disease and inferolateral myocardial infarction. Diastolic (A–C) and systolic (D–F) CT image reformations show reduced diastolic wall thickness and absence of systolic wall thickening of inferior and inferolateral walls of left ventricle. Left ventricular volumes and ejection fraction determined with 3D threshold segmentation algorithm (A, B, D, E) and 2D short axis–based planimetry (Simpson method) (C, F) show comparable results.

 

One reason for the higher success rate with the 3D segmentation algorithm in our study may be that we used a different protocol for contrast administration. Dual-source CT coronary angiography was performed with a larger initial bolus (70 mL) of contrast material at a higher flow rate (5 mL/s) and was followed by a diluted (40-mL) second bolus and a saline chaser. In all patients, this CT protocol resulted in a greater than 150-H difference in attenuation values in the LV lumen compared with the intraventricular myocardial septum. This cutoff value enabling success of the automated 3D segmentation technique had been recommended [16].

We found excellent correlation between the findings with threshold-based segmentation and those with short-axis planimetry for LVEDV and LVESV, though there was moderate systematic underestimation of LV volumes. We believe the systematic underestimation of LV volumes with threshold-based segmentation is explained by different handling of the papillary muscles. With attenuation-controlled segmentation, structures with soft-tissue density, such as the papillary muscles, are excluded from the LV volume, whereas standard short axis–based LV functional analysis usually includes these structures as part of the LV cavity [7–14, 18–28]. Vogel-Claussen and coauthors [29] found that papillary muscle volume accounted for 10.5 ± 2.5% of the LV blood pool volume, and van der Geest et al. [30] reported that the papillary muscles represent 6.5% of the LVEDV on short-axis planimetry.

We found slight (2.8%) systematic overestimation of LVEF with threshold-based 3D segmentation. Again, this difference was most likely due to the difference in measurement approach, but it may reflect bias in determination of the most basal slice of the left ventricle. Because of the large area of the base of the heart, inclusion or exclusion of a basal slice results in considerable variation in LV volume. From a clinical perspective, however, 3% deviation in LVEF can be regarded as rather insignificant, especially in light of the measurement variations reported for cine MRI and echocardiography.

With threshold-based 3D segmentation, we found excellent interobserver agreement for all LV volumes and a negligible 0.1% difference in LVEF. We believe this very low interobserver variability, especially in light of the limited experience of reader 2, constitutes a major advantage of threshold-based 3D segmentation compared with the traditional 2D Simpson method, which had an interobserver variation of 4.3% for LVEF (p < 0.01). There are only limited data regarding interobserver variability in LV functional analysis MDCT. Interobserver variability of 0.5–11% has been reported for LVEDV and LVESV and of 2–8.5% for LVEF [8, 11, 21, 24, 27]. The most likely explanation for the considerably higher interobserver variation is subjective determination of the most basal and apical slices and manual or semiautomatic endocardial contour definition. In contrast to the 2D short-axis approach, use of axial slices in the 3D approach leads to easier and thus more reproducible detection of the mitral valve plane owing to better delineation of the left atrium versus the left ventricle. Because of the small interobserver variability and the fact that data analysis was performed by two readers with differing levels of expertise in cardiac radiology (staff radiologist vs first-year resident), threshold-based 3D segmentation may prove advantageous for follow-up assessment of LV function with CT coronary angiographic studies.

Dual-source CT provides combined morphologic and functional clinical data on patients undergoing CT coronary angiography, although it has not been considered a first-line technique for assessment of global and regional LV function. However, the technique may develop as an alternative diagnostic procedure for acquiring additional functional information about patients with implanted pacers and cardioverter–defibrillators, especially patients who have undergone cardiac resynchronization therapy. MDCT regional LV wall motion analysis at rest is feasible, but the sensitivity for detection and accurate classification of LV wall motion abnormalities must be improved [28]. Further improvement of the temporal resolution of the MDCT system is crucial to match results of regional LV functional analysis obtained with echocardiography and cine MRI. With significantly improved temporal resolution, even better agreement between quantitative MDCT and cine MRI assessments of regional LV function is to be expected with dual-source CT systems.

Our study was focused on clinical evaluation of the threshold-based 3D segmentation software (n = 50) in comparison with the established 2D short axis–based approach to state-of-the art dual-source CT technology. Validation of LV functional assessment with dual-source CT against cine MRI was not intended. Because accurate LV volume measurements depend on the temporal resolution of the imaging technique used [31, 32] and use of MDCT has been shown to yield underestimates of LVESV compared with use of cine MRI [33, 34], improvement in LV functional measurements has to be expected with use of dual-source CT technology that addresses this shortcoming by providing a temporal resolution of 83 milliseconds independent of heart rate.

Although patients with implanted pacers and implanted cardioverter–defibrillators were not systematically excluded from this study, none of the patients had metallic right or left ventricular implants. In principle, the presence of pacer or cardioverter–defibrillator leads can cause artifacts that impair proper function of the threshold-based segmentation algorithm. Systematic evaluation of this specific group of patients will be of further interest.

Automated threshold-based 3D segmentation allows accurate and reproducible estimation of LV volumes and function from dual-source CT coronary angiographic data sets in close correlation with results obtained with the established 2D short-axis approach. Exclusion of papillary muscles from the LV volume results in small systematic differences in quantitative values.


References
Top
 Abstract
 Introduction
 Subjects and Methods
 Results
 Discussion
 References
 

  1. Hammermeister KE, DeRouen TA, Dodge HT. Variables predictive of survival in patients with coronary disease: selection by univariate and multivariate analyses from the clinical, electrocardiographic, exercise, arteriographic, and quantitative angiographic evaluations. Circulation 1979;59 : 421–430[Abstract/Free Full Text]
  2. Shah PK, Maddahi J, Staniloff HM, et al. Variable spectrum and prognostic implications of left and right ventricular ejection fraction in patients with and without clinical heart failure after acute myocardial infarction. Am J Cardiol 1986;58 : 387–393[CrossRef][Medline]
  3. White HD, Norris RM, Brown MA, Brandt PW, Whitlock RM, Wild CJ. Left ventricular end-systolic volume as the major determination of survival after recovery from myocardial infarction. Circulation1987; 76:44 –51[Abstract/Free Full Text]
  4. Desjardins B, Kazerooni EA. ECG-gated cardiac CT. AJR 2004; 182:993 –1010[Free Full Text]
  5. Achenbach S. Computed tomography coronary angiography. J Am Coll Cardiol 2006; 48:1919 –1928[Abstract/Free Full Text]
  6. Orakzai SH, Orakzai RH, Nasir K, Budoff MI. Assessment of cardiac function using multidetector row computed tomography. J Comput Assist Tomogr 2006; 30:555 –563[CrossRef][Medline]
  7. Schlosser T, Pagonidis K, Herborn CU, et al. Assessment of left ventricular parameters using 16-MDCT and new software for endocardial and epicardial border delineation. AJR 2005;184 : 765–773[Abstract/Free Full Text]
  8. Yamamuro M, Tadamura E, Kubo S, et al. Cardiac functional analysis with multi-detector row CT and segmental reconstruction algorithm: comparison with echocardiography, SPECT, and MR imaging. Radiology 2005;234 : 381–390[Abstract/Free Full Text]
  9. Juergens KU, Grude M, Maintz D, et al. Multi-detector row CT of left ventricular function with dedicated analysis software versus MR imaging: initial experience. Radiology 2004;230 : 403–410[Abstract/Free Full Text]
  10. Mahnken AH, Mühlenbruch G, Koos R, et al. Automated vs. manual assessment of left ventricular function in cardiac multidetector row computed tomography: comparison with magnetic resonance imaging. Eur Radiol 2006; 16:1416 –1423[CrossRef][Medline]
  11. Belge B, Coche E, Pasquet A, Vanoverschelde JL, Gerber BL. Accurate estimation of global and regional cardiac function by retrospectively gated multidetector row computed tomography: comparison with cine magnetic resonance imaging. Eur Radiol 2006;16 :1424 –1433[CrossRef][Medline]
  12. Dewey M, Muller M, Teige F, et al. Multisegment and halfscan reconstruction of 16-slice computed tomography for assessment of regional and global left ventricular myocardial function. Invest Radiol 2006; 41:400 –409[CrossRef][Medline]
  13. Boehm T, Alkadhi H, Roffi M, et al. Time-effectiveness, observer-dependence, and accuracy of measurements of left ventricular ejection fraction using 4-channel MDCT. Rofo 2004;176 : 529–537[Medline]
  14. Juergens KU, Seifarth H, Maintz D, et al. Left ventricular function determination with multidetector-row CT of the heart: are short-axis image reformations necessary? AJR 2006;186 :S371 –S378[Abstract/Free Full Text]
  15. Erhard K, Oberholzer K, Gast K, Mildenberger P, Kreitner KF, Thelen M. Multi-slice CT (MSCT) in cardiac function imaging: threshold-value supported 3D volume reconstructions to determine the left ventricular ejection fraction in comparison to MRI. Rofo 2002;174 :1566 –1569[Medline]
  16. Mühlenbruch G, Das M, Hohl C, et al. Global left ventricular function in cardiac CT: evaluation of an automated 3D region-growing segmentation algorithm. Eur Radiol 2006;16 :1117 –1123[CrossRef][Medline]
  17. Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet1986; 1:307 –310[CrossRef][Medline]
  18. Dirksen MS, Bax JJ, de Roos A, et al. Usefulness of dynamic multislice computed tomography of left ventricular function in unstable angina pectoris and comparison with echocardiography. Am J Cardiol 2002; 90:1157 –1160[CrossRef][Medline]
  19. Juergens KU, Grude M, Fallenberg EM, et al. Using ECG-gated multidetector CT to evaluate global left ventricular myocardial function in patients with coronary artery disease. AJR2002; 179:1545 –1550[Abstract/Free Full Text]
  20. Schuijf JD, Bax JJ, Salm LP, et al. Noninvasive coronary imaging and assessment of left ventricular function using 16-slice computed tomography. Am J Cardiol 2005;95 : 571–574[CrossRef][Medline]
  21. Grude M, Juergens KU, Wichter T, et al. Evaluation of global left ventricular myocardial function with electrocardiogram-gated multidetector computed tomography: comparison with magnetic resonance imaging. Invest Radiol 2003;38 : 653–661[Medline]
  22. Lessick J, Mutlak D, Rispler S, et al. Comparison of multidetector computed tomography versus echocardiography for assessing regional left ventricular function. Am J Cardiol 2005;96 :1011 –1015[CrossRef][Medline]
  23. Heuschmid M, Rothfuss JK, Schroeder S, et al. Assessment of left ventricular myocardial function using 16-slice multidetector-row computed tomography: comparison with magnetic resonance imaging and echocardiography. Eur Radiol 2006;16 : 551–559[CrossRef][Medline]
  24. Sugeng L, Mor-Avi V, Weinert L, Quantitative assessment of left ventricular size and function: side-by-side comparison of real-time threedimensional echocardiography and computed tomography with magnetic resonance reference. Circulation 2006;114 : 654–661[Abstract/Free Full Text]
  25. Dewey M, Muller M, Eddicks E, et al. Evaluation of global and regional left ventricular function with 16-slice computed tomography, biplane cineventriculography, and two-dimensional transthoracic echocardiography. J Am Coll Cardiol 2006;48 :2034 –2044[Abstract/Free Full Text]
  26. Henneman MM, Schuijf JD, Jukema JW, et al. Assessment of global and regional left ventricular function and volumes with 64-slice MSCT: a comparison with 2D echocardiography. J Nucl Cardiol2006; 13:480 –487[CrossRef][Medline]
  27. Schepis T, Gaemperli O, Koepfli P, et al. Comparison of 64-slice CT with gated SPECT for evaluation of left ventricular function. J Nucl Med 2006; 47:1288 –1294[Abstract/Free Full Text]
  28. Fischbach R, Juergens KU, Ozgun M, et al. Assessment of regional left ventricular function with multidetector-row computed tomography versus magnetic resonance imaging. Eur Radiol2007; 17:1009 –1017[CrossRef][Medline]
  29. Vogel-Claussen J, Finn JP, Gomes AS, et al. Left ventricular papillary muscle mass: relationship to left ventricular mass and volumes by magnetic resonance imaging. J Comput Assist Tomogr2006; 30:426 –432[CrossRef][Medline]
  30. van der Geest RJ, Buller VG, Jancen E, et al. Comparison between manual and semiautomated analysis of left ventricular volume parameters from short axis MR images. J Comput Assist Tomogr1997; 21:756 –765[CrossRef][Medline]
  31. Miller S, Simonetti OP, Carr J, et al. MR imaging of the heart with cine true fast imaging with steady-state precession: influence of spatial and temporal resolution on left ventricular functional parameters. Radiology 2002;223 : 263–269[Abstract/Free Full Text]
  32. Mahnken AH, Hohl C, Suess C, et al. Influence of heart rate and temporal resolution on left-ventricular volumes in cardiac multislice spiral computed tomography: a phantom study. Invest Radiol2006; 41:429 –435[CrossRef][Medline]
  33. Flohr TG, McCollough CH, Bruder H, et al. First performance evaluation of a dual-source CT (DSCT) system. Eur Radiol 2006; 16:235 –268
  34. Rist C, Johnson TR, Becker A, et al. Dual-source cardiac CT imaging with improved temporal resolution: impact on image quality and analysis of left ventricular function. Radiologe2007; 47:287 –294[CrossRef][Medline]

Add to CiteULike CiteULike   Add to Complore Complore   Add to Connotea Connotea   Add to Del.icio.us Del.icio.us   Add to Digg Digg   Add to Reddit Reddit   Add to Technorati Technorati    What's this?



This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF)
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Right arrow Citation Map
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Juergens, K. U.
Right arrow Articles by Fischbach, R.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Juergens, K. U.
Right arrow Articles by Fischbach, R.
Social Bookmarking
Add to CiteULike   Add to Complore   Add to Connotea   Add to Del.icio.us   Add to Digg   Add to Reddit   Add to Technorati  
What's this?
Hotlight (NEW!)
Right arrow
What's Hotlight?

HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS