| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Original Research |
1 Department of Diagnostic Radiology, Eberhard-Karls-University,
Hoppe-Seyler-Str. 3, 72076 Tübingen, Germany.
2 Department of Cardiology, Eberhard-Karls-University, Tübingen,
Germany.
Received March 11, 2007;
accepted after revision June 5, 2007.
Address correspondence to H. Brodoefel
(h.brodoefel{at}t-online.de).
Abstract
 Top Abstract IntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
SUBJECTS AND METHODS. Twenty patients (15 of whom had previous myocardial infarction) were prospectively examined using dual-source CT. MRI was used as the standard of reference. Using the Simpson's method, ventricular volumes were determined for the whole of the cardiac cycle and results compared using Parson's correlation and Bland-Altman analysis. Regional wall motion was assessed on cine images and compared using weighted kappa statistics.
RESULTS. Dual-source CT revealed a strong correlation with cine MRI
regarding the quantification of end-diastolic volume (r = 0.98),
end-systolic volume (r = 0.99), stroke volume (r = 0.96),
and ejection fraction (r = 0.95). Good correlation was obtained for
peak ejection rate (r = 0.79) and peak filling rate (r =
0.84), whereas agreement proved only moderate for time-to-peak ejection rate
(r = 0.68) or time-to-peak filling rate from end-systole (r
= 0.64). The mean difference for ejection fraction was negligible (bias,
0.72%). Good agreement between both techniques was likewise found for regional
wall motion (
= 0.88).
CONCLUSION. With the improvement of temporal resolution between 42 and 83 milliseconds, dual-source CT not only enables accurate assessment of global functional parameters, but it also allows for quantification of time-dependent variables and reliable evaluation of regional wall motion.
Keywords: dual-source CT • left ventricular function • MRI
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
Because of the considerable radiation exposure, MDCT has seldom been used for the analysis of LV function. Nevertheless, in the past few years, technical progression from 4- to 64-MDCT systems has opened the field of noninvasive coronary angiography, leading to a widespread acceptance and use of MDCT as a noninvasive technique for cardiac imaging [5, 6]. Because information on any conceivable heart phase is always obtained as a by-product of coronary imaging and may easily be extracted from available data sets, the role of CT for functional analysis has ultimately grown.
In the past, a number of investigations have proven a close correlation between cine MRI and CT estimation of such global functional parameters as end-diastolic volume (EDV), end-systolic volume (ESV), stroke volume (SV), and ejection fraction (EF) [7–12]. However, due to a limited temporal resolution of previous MDCT generations, the method has proven unsatisfactory in the assessment of time-dependent parameters such as peak filling rate (PFR), peak ejection rate (PER), time to PER, and time from end-systole (ES) to PFR [13, 14]. For the same reasons, CT analysis of regional wall motion has previously been reported to be inadequate [10].
However, recently, a new CT system equipped with two tubes and corresponding detectors in a 90° geometry has been designed and provides a heart-rate-independent temporal resolution of 83 milliseconds. The latter may be further improved by using a multisegment reconstruction algorithm and thus be as low as 42 milliseconds [15]. The doubling of temporal resolution compared with the latest CT generations with 330 milliseconds gantry rotation time holds the promise of ultimately eliminating many of the limitations that have previously been faced in CT functional analysis.
Hence, the aim of our study was to assess the performance of increased temporal resolution in the determination of both volume- or time-dependent functional parameters and regional wall motion in comparison with cine MRI.
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
Dual-Source CT Scanning Protocol
All CT examinations were performed on a dual-source CT system (Somatom
Definition, Siemens Medical Solutions). Scanning parameters were as follows:
120 kV, 550 mAs for each X-ray tube, 0.33-millisecond gantry rotation, 32
x 0.6 mm collimation with z-flying focal spot for each
detector, and pitch of 0.20. A dual-segment reconstruction algorithm was
applied to all data, regardless of patient heart rate. Temporal resolution was
thereby varying as a function of the patient's heart rate between 83 and 42
milliseconds. Using dual-source CT in the dual-segment reconstruction mode, a
mean resolution of 60 milliseconds can be established
[15]. For dose reduction, the
automatic tube current modulation device was used.
The scanning delay was determined using the test bolus technique with the region of interest placed in the ascending aorta. A test bolus of 20 mL of iomeprol (Imeron 400, Bracco) was followed by a 40-mL saline chaser bolus. For CT examination of the heart, 100 mL of contrast material was injected IV at 4–5 mL/s and followed by a 50-mL saline chaser bolus. Compared with routine noninvasive coronary angiography, the amount of contrast material was increased by 20 mL to ensure additional opacification of the right ventricle.
After the acquisition of the 3D data sets, about 20 series of axial images were reconstructed every 5% (0–95%) of the R-R interval with an effective slice thickness of 0.75 mm, a reconstruction increment of 0.4 mm, and a field of view of 180 x 170 mm (matrix, 512 x 512). For further analysis, 5-mm multiplanar reformations without an intersection gap were calculated along the short axis and in the two-, three-, and four-chamber orientation.
MRI
MRI was performed on a 1.5-T whole-body MR scanner (Magnetom Sonata,
Siemens Medical Solutions) using a phased-array body coil. After obtaining
localizing scout images, breath-hold true fast imaging with steady-state
precession (FISP) cine sequences with retrospective ECG gating were acquired
with TR/TE, 3.08/1.54; flip angle, 55°; matrix, 256 x 208; and pixel
size, 1.35 x 1.35 mm. Neither generalized autocalibrating partially
parallel acquisitions (GRAPPA) nor sensitivity encoding (SENSE) was used.
About 20 phase-encoding steps were used per time frame, with a resulting
temporal resolution of 30–50 milliseconds, depending on heart rate.
Images with a section thickness of 5 mm and an intersection gap of 5 mm were
obtained in standardized short-axis and two-, three-, or four-chamber
orientation.
LV Function Analysis
LV function analysis was performed on an offline workstation (Leonardo,
Siemens Medical Solutions) using commercially available software (Argus,
Siemens Medical Solutions). CT and MR images were separately assessed by two
experienced radiologists with 5 and 8 years of experience who were blinded to
the results of the other imaging technique. Window settings were individually
adapted by the readers using the half-contour principle
[16].
Global LV function was assessed on short-axis images using the automated endocardial or epicardial contour detection function with subsequent visual control and manual correction. The most basal slice surrounded by at least 50% myocardium was defined as the LV base, whereas the first slice with a visible lumen during the entire cardiac cycle was chosen as the apex of the LV. The papillary muscles were included in the ventricular volume. End-diastole and end-systole were defined as maximal and minimal LV volume, and all functional parameters were calculated according to the modified Simpson's method [17].
Segmental wall motion analysis was performed according to a 17-segment model of the LV [18]. Regional wall motion was assessed from cine loops in the short axis and in the two-, three-, and four-axis orientation. Motion was visually evaluated on a per-segment basis and described as normal, hypokinetic, akinetic, or dyskinetic.
Intra- and Interobserver Variability of CT Measurements
For assessment of intraobserver variability, a second reading was performed
on all CT data after a 4-month interlude. Interobserver variability was
analyzed by comparing CT measurements of two separate investigators, each with
5 years of experience.
Statistical Analysis
All continuous variables are expressed as mean ± SD, and comparison
between methods was performed using the Student's t test for paired
observations. Bland-Altman analysis was used to display the systematic error
and the limits of agreement between the two methods. The correlation between
parameters was assessed by calculation of the Pearson's correlation
coefficient. Results of regional wall motion assessment were displayed in a
contingency table, and agreement between CT and MRI was analyzed by use of
weighted kappa statistics. Inter- or intraobserver reliability was assessed by
calculating two-way random or one-way random intraclass correlation
coefficients (ICCs). For all statistical analysis, a p value of less
than 0.05 was considered significant. All analysis was performed with SPSS,
version 15, or GraphPad Prism, version 4.00.
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
|
|
|
|
Results of volume and function measurements are summarized in Table 1. There was a high correlation of dual-source CT and MRI values for EDV, ESV, EF, and SV. Strong association was also obtained for such time-dependent variables as PER or PFR, whereas there was only moderate agreement for time to PER or time to PFR from ES. The agreement of time-dependent parameters is mirrored in the uniform pattern of time–volume curves obtained by both techniques (Fig. 2A, 2B).
|
|
|
Comparing dual-source CT with MRI, Bland-Altman analysis revealed a mean bias for EDV of –2.2 mL, for ESV of –1.4 mL, for EF of 0.7%, for SV of –2.5 mL, for PER of –17.3 mL/s, for PFR of –26.9 mL/s, for time to PER of –7.7 milliseconds, and for time from ES to PFR of –11.0 milliseconds (Figs. 3A, 3B, 3C, 3D and 4A, 4B, 4C, 4D). Thereby, only overestimation of EDV proved significantly different (p = 0.042). In addition, there was a considerable trend toward overestimation of PER and PFR (p = 0.12 and p = 0.069, respectively).
|
|
|
|
|
|
|
|
|
Inter- and intraobserver reliability proved high for LV volumes and good for time-dependent variables (Table 3). According to CT reader 2, there was no significant trend toward overestimation of EDV in dual-source CT (p = 0.21).
|
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
Our primary finding is that dual-source CT with enhanced temporal resolution not only provides good correlation for global functional parameters such as EF or SV but also accurately determines time-dependent parameters such as PER or PFR. Good agreement with cine MRI was likewise achieved for the assessment of regional wall motion.
The general capacity of MDCT to assess such global functional parameters as peak systolic or diastolic ventricular volume has previously been established by several investigators in 4- or 16-MDCT [7–12, 14, 19]. However, as is known from cardiac MRI, temporal resolution has a crucial impact on accuracy, and given the shortness of isovolumetric relaxation of the end-systole (40–60 milliseconds), a temporal resolution of 50 milliseconds or less has been defined as optimal for exact quantification [20, 21].
Whereas with 4-MDCT scanners, limited temporal resolution and the failure to capture the phase of maximum systolic contraction still led to systematic overestimation of ESV, the application of 16-MDCT allowed robust quantification of ventricular volumes with high correlation to cine MRI. In a pooled analysis of EF measurements from 8- and 16-MDCT scanners, the overall weighted average bias was found to be–1.8% [22]. For 64-MDCT systems with further improved temporal resolution, head-to-head comparisons with MRI are not yet available; however, promising results have been published with 2D echocardiography or gated SPECT used as the standard of reference [23, 24].
In our study, a multisegment reconstruction algorithm has been applied with temporal resolution, thus varying with heart rate between 42 and 83 milliseconds. Mean resolution was 60 milliseconds. Regarding global functional parameters, we found a slight and clinically irrelevant overestimation of EDV and ESV, whereas relative measurements such as EF were in excellent agreement, with a systematic error of less than 1%. Regarding the satisfactory correlation of time-dependent parameters in our study, differences between EDV or ESV were most likely not due to an inferior temporal resolution in dual-source CT. Interestingly, a meta-analysis of recent comparisons between MDCT and MRI has proven a small but systematic overestimation for ventricular volumes in MDCT [22]. A plausible explanation would be that a different level of contrast between blood and myocardium causes divergence in delineation of the trabecular–papillary muscle complex. The changing configuration of this complex during various heart cycle phases may likewise explain isolated overestimation of either EDV or ESV, as was observed in our study.
Reliable estimation of such global functional parameters as EDV, ESV, and EF is of the essence because they are independent predictors of morbidity and mortality in patients with CAD [25]. Notably, these variables are determined from LV contours at only two points of the cardiac cycle. Nevertheless, it has been shown that time–volume curves representing continuous LV volume changes throughout the cardiac cycle allow a more detailed quantitative analysis of LV performance [26–28]. In fact, such parameters as PER or PFR provide a very sensitive and early detection of ejection or filling abnormalities and thereby allow a differentiation of either isolated systolic or diastolic dysfunction or a combination of both. Through the use of time–volume curves, progression of disease or response to therapy may thus more accurately be followed [29–32]. Due to insufficient temporal resolution of previous CT scanners, however, assessment of time-dependent functional parameters has proven inaccurate and could not be used in clinical routine [13, 14].
Our results confirm that the improvement in temporal resolution translates into the feasibility of reliably quantifying time-dependent parameters, which show good to moderate agreement with measurements from MRI. Thereby, a systematic overestimation was noted for all time-dependent parameters obtained, and differences were almost significant for PER and PFR. Correlation with cine MRI was only moderate for time to PER or time from ES to PFR. For all time-dependent variables, a considerable scatter of values in both methods was causative of relatively poor limits of agreement.
In our study, intra- and interobserver reliability for time-dependent variables was not as good as for single ventricular volumes, notably EDV or ESV. This reflects the complex nature of parameters such as time to PER or PFR, which are calculated from multiple volumes within the cardiac cycle. However, the reliability of measurements was still high, and therefore only a small part of scatter may be explained by interobserver variability. Ultimately, the limited temporal resolution of dual-source CT remains the primary cause of only moderate correlation with cine MRI. Nevertheless, the general capacity to obtain time-dependent functional parameters with temporal resolution of 60 milliseconds is a new and interesting feature and certainly an asset for functional analysis using dual-source CT.
Limited temporal resolution of MDCT has likewise been a limitation in the evaluation of regional wall motion, which generally necessitates a temporal resolution of 80 milliseconds at rest—a requirement not met by previous CT generations [21]. Nevertheless, in the past, several investigators haven proven that assessment of regional wall motion is generally feasible and that moderate or good correlation may be obtained. With 16-MDCT and in comparison with MRI, the level of segmental agreement is reported to vary between 86% and 98% [10, 14, 19]. Although the 64-MDCT system has not yet been compared with MRI, a 96% agreement in regional wall motion has recently been reported in comparison with 2D echocardiography [23].
In our study, there was a 96.7% segmental agreement between MRI and dual-source CT regarding the assessment of regional wall motion. However, ratings remained discrepant in 3.3% of all included segments or 22.5% of only dysfunctional segments. Interestingly, the dominant limitation of dual-source CT was found to be the use of tube current modulation, with the periodic decrease of tube current and subsequent increase of image noise in systolic heart phases. In fact, although poor contrast opacification or conceivable motion artifacts could be excluded for all of our patients, the degradation of image quality in systole was retrospectively identified as the single reason for discrepant ratings in our study. Of note, in most of the previous studies on regional wall motion, tube current modulation was not used.
Nevertheless, modulation of tube current has been proven to be an effective tool for dose reduction and may limit patient dose by up to 47% according to the individual ECG [33]. Hence, omission of such a technical device seems not to be justified. Indeed, further control studies are warranted to exactly quantify the loss of accuracy with greater systolic image noise; the window width and lowest current of modulation may then be adjusted to special clinical issues.
In this context, it is important to note that modulation of tube current in our study did not impede correct identification of epicardial or endocardial contours on axial images. On the contrary, because of clearer delineation of myocardial margins in dual-source CT, the use of the semiautomated contour setting function proved more successful in CT and required fewer corrections when compared with MRI.
In noninvasive coronary angiography, ß-blockers are widely used to
lower the heart rate and thereby reduce motion artifacts of coronary arteries.
However, the artificial reduction of heart rate may alter functional
parameters and may thus not reflect true cardiac performance
[34]. Therefore, no additional
ß-blocker medication was administered in our study. Of note, even in
patients with heart rates 
75 bpm, motion artifacts were absent and LV
contours could be well defined.
The small number of observations and the considerable heterogeneity of underlying disease and degree of ventricular dysfunction is an obvious limitation to our study. Bias and SD for time-dependent parameters need to be confirmed in larger studies, and the evaluation of characteristic time–volume curve patterns for various types of cardiac insufficiencies is warranted.
In the present study, a slice gap of 5 mm was used to achieve a moderate examination time on MRI; by contrast, dual-source CT data sets were based on gapless 5-mm sections. Nevertheless, because temporal resolution is far more important for accurate evaluation of advanced functional parameters, this discrepancy of method is unlikely to cause significant distortion of results [20]. Opposing breath-holding techniques with resulting discrepancy of intrathoracic pressure may be a potential cause of systematic bias. Finally, susceptibility of ECG signal in MRI is a general source of discrete random error.
A serious limitation to routine use of advanced or time-dependent functional parameters is the impracticability of postprocessing. In our study, contour setting for multiple slices in every phase of the cardiac cycle required about 25–30 minutes for each technique. This was despite automatic contour setting and because of the many observer interactions required. Such unacceptable time constraints must be overcome by more sophisticated and automated postprocessing techniques before time–volume curves become a routine mosaic of LV functional assessment.
In conclusion, improved temporal resolution of dual-source CT not only provides an accurate analysis of global LV function but for the first time allows additional evaluation of time-dependent variables, even though their agreement with MRI is only moderate. Regional wall motion is assessed with good accuracy. With the increase of temporal resolution below 83 milliseconds, many of the previous limitations thus seem to have been overcome, and MDCT eventually will evolve as an attractive tool for combined evaluation of coronary arteries and ventricular function.
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
