| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Original Research |
1 Department of Radiology, University Hospitals of Cleveland, 11100 Euclid Ave.,
Cleveland, OH 44106.
2 Department of Radiology, University Hospitals of Ulm, Ulm, Germany.
Received November 17, 2004;
accepted after revision March 3, 2005.
Address correspondence to D. T. Boll
(boll{at}uhrad.com).
Abstract
 Top Abstract IntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
SUBJECTS AND METHODS. Twenty patients underwent concurrent ECG-gated MDCT and MRI for evaluation of ventricular motility, expressed as ventricular wall thickening and motion. Initially, individual measurement repetition series were analyzed by determining intraobserver variability and detecting intraobserver bias related to heart rates. Subsequently, absolute measurement differences of CT or MR data were statistically evaluated. Finally, absolute measurement differences were correlated with underlying heart rates by curve estimation regression.
RESULTS. Analysis of measurement reproducibility proved that data variability was dependent on only the anatomic localization of the analyzed ventricular segment, not on the imaging technique used or underlying heart rate, in normofrequent patients. Comparing MR and CT image data sets, no statistically significant differences were identified when ventricular motility was evaluated based on data sets of either imaging technique in normofrequent patients. Tachycardic frequencies, above 100 beats per minute, led to exponential error propagation due to insufficient temporal resolution of the current CT technology.
CONCLUSION. This study proved that cardiac motility assessment based on ECG-gated CT and MR data sets resulted in comparable ventricular function results for normofrequent patients; however, the high spatial resolution of cardiac MDCT cannot compensate for the lack of temporal resolution in patients with tachycardia, thus emphasizing the necessity of reporting ventricular motility analysis results in combination with heart rate to allow consideration of this possible cause for measurement variation.
Keywords: cardiac imaging • cine MRI • MDCT
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
More recently, cross-sectional imaging techniques with ECG gating such as MDCT and cine MRI have been used increasingly in the assessment of dynamic cardiac function. ECG-gated CT and MRI have the potential to provide high-resolution image series of the heart at any point in the cardiac cycle. In particular, because cardiac MRI can provide excellent contrast of the myocardium and ventricular blood pool and sufficient spatial resolution to differentiate between papillary muscles and myocardium, many consider it to be the gold standard for dynamic cardiac function assessment [5-7]. With the introduction of ECG-gated MDCT technology, yet another technique with the potential to analyze cardiac function became available [8, 9].
Reproducible and interchangeable functional analysis based on cardiac MRI and CT examinations is crucial for the complementary and synergistic application of these imaging techniques. This study was performed to test the hypothesis that successful and accurate assessment of ventricular motility is independent of either source of imaging data and that these imaging techniques, with their specific abilities to depict various individual cardiac conditions, can be used interchangeably for the assessment of ventricular motility.
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
The subsequent contrast-enhanced ECG-gated cardiac CT protocol was performed, first, to exclude underlying cardiomyopathy—in particular, the noncompaction type; second, to analyze myocardial motility and mechanical heart valve integrity by calculating 10 equally spaced cardiac phases in the short-axis orientation at 2-mm slice thickness of the raw data set; and, third, to assess the pericardium—in particular, for the presence of constrictive pericarditis—using late systolic and late diastolic phases at 2-mm slice thickness. Furthermore, coronary CT angiography functioned as a gatekeeper for potential conventional coronary angiography with possible intervention by revealing the patency of the coronary arteries and coronary stentgrafts, allowing the degree of stenosis of individual coronary arteries to be estimated, and depicting coronary anomalies by calculating 0.6-mm slices from the same raw data set with subsequent automated coronary centerline extraction and evaluation. Neither adenosine stress MRI nor dobutamine stress echocardiography was part of the standardized initial cardiac workup protocol; whether to perform subsequent additional stress imaging was decided on the basis of the results of the initial cardiac workup.
Twenty patients—that is, seven women and 13 men—were referred to an institution for an initial cardiac workup during the period from October 2002 to September 2003; none had undergone any prior imaging, so the cardiac workup protocol was modified. These 20 patients were the original study population of this prospective study. The women ranged in age from 18 to 86 years with a mean age of 53.1 ± 24.9 (SD) years, and the men averaged 56.3 ± 18.6 years, ranging in age from 29 to 84 years. Exclusion criteria for the comprehensive cardiac workup and this study were an allergy to iodine-containing contrast media; renal failure (serum creatinine > 100 µmol/L); pregnancy; severe respiratory impairment or pronounced cardiac failure; and implanted cardiac pacemakers, defibrillators, or any other incorporated ferromagnetic objects. Subsequent patient recruitment was consecutive without exclusion. Informed consent covering the use of raw data sets for this study was obtained from each patient. The study group size of 20 patients corresponds to a power of 0.953, being able to reliably detect a difference of 0.2 from the null hypothesized proportion of 0.1 using a normal approximation power analysis.
The cardiac conditions of the study population included dilated cardiomyopathy (n = 4); aortic valve disease (n = 4); constrictive pericarditis (n = 7); cardiac masses, such as atrial (n = 1) and ventricular (n = 1) myxoma; and various congenital heart diseases, such as noncompaction cardiomyopathy (n = 2); and an anomalous left ventricular outflow tract (n = 1).
Image Acquisition
Cine MRI and MDCT examinations were performed sequentially in this order on
the same day with an average intermission of 2.4 ± 0.3 (SD) hr (range,
1-4 hr) in all patients. In all cases, MR-compatible ECG electrodes were
placed on the chest before imaging and were attached to the CT and the MR
imagers for ECG gating. Patient heart rates ranged from 68 to 123 beats per
minute (bpm) with a mean rate of 97.3 ± 15.5 bpm and an average
variation of 6.1 bpm during the examinations.
All MRI examinations were performed on a 1.5-T imaging system (Magnetom Sonata, Siemens Medical Solutions) equipped with high-performance gradients with a maximum amplitude of 40 mT/m and a maximum slew rate of 200 mT/m/msec using a 4-channel torso phased-array coil. Two long-axis and short-axis breath-hold cine image series sufficient to cover the whole left ventricle were acquired using a 5-mm slice thickness and a 2.5-mm gap. Fast imaging without additional contrast application was used with steady-state precession with balanced gradients in all spatial directions (true FISP) in all patients. The TR/TE was 2.96/1.58; a flip angle of 70° was chosen with the number of signal averages being 1, and the matrix size was set to 208 x 256 pixels, resulting in an in-plane spatial resolution of 1.38 mm and a voxel volume of 9.52 mm3. The acquisition time per cine frame and heartbeat was 47.5 msec; however, 15 heartbeats were necessary in this prospectively cardiac-gated sequence for a complete interleaved filling of k-space; therefore, the total acquisition time of 15 heartbeats was approximately 12-15 sec per each long-axis and short-axis location. A cartesian rectilinear k-space trajectory was used with a central reordering scheme [10].
The subsequent CT examinations were performed on an MDCT scanner (Mx8000IDT, Philips Medical Systems) characterized by a parallel arrangement of 16 detector arrays. All studies were preceded by a scout acquisition and were performed during a single breath-hold with the patient in the supine position. A helical CT study was planned to cover a longitudinal field of view from the diaphragm to the carina. The helical scanning protocol consisted of a detector pitch of 0.24, a gantry rotation period of 0.42 sec, a collimation of 16 x 0.75 mm, a table speed of 6.8 mm/sec, and a matrix size of 512 x 512 pixels applying an X-ray tube voltage of 140 kV. To reduce radiation exposure to the patient, an X-ray tube current of 550 mA adjusted by adaptive dose modulation techniques was applied resulting in an effective patient dose range of 7.2-9.1 mSv depending on the scan length for the entire examination.
For automatic bolus tracking, axial slices were acquired at the level of the left ventricle every 1.7 sec at 6-mm slice thickness using low-dose X-ray tube parameters such as 120-kV voltage and 30-mA current. A region of interest (ROI) with an average diameter of 12.4 ± 5.1 mm was positioned in the left ventricle and enhancement (in Hounsfield units) was plotted against time. Uniformly, 100 mL of low-osmolar iodinated contrast agent (ioversol [Optiray 300, Mallinckrodt]) was administered via an 18-gauge IV line placed in an antecubital vein with a 4.0 mL/sec flow rate followed by a 40-mL saline chaser bolus with a 3.5 mL/sec flow rate. The helical scan was automatically initiated with a delay of 8 sec according to the peak enhancement derived from the ROI above a threshold level of 150 H. The duration of the entire examination was 20-29 sec.
After acquisition of the raw helical CT data, retrospective ECG-synchronized slices were reconstructed. A dedicated reconstruction algorithm for ECG-gated MDCT was used by combining partial scan reconstructions optimized for temporal resolution with low-pitch multislice spiral weighting covering the heart with longitudinal redundancy over multiple heart cycles. Therefore, temporal resolution and spatial resolution were dependent on the pitch setting and the collimation in combination with the underlying heart rate [11]. In particular, single acquisition frame times, based on the pitch setting and varying heart rates, presented with individual, unpredictable variations due to complex relationships between synchronously and asynchronously moving structures, the CT gantry rotation time and table speed, and the heart. At heart rates above 65 bpm, raw data were combined from multiple consecutive cardiac cycles, resulting in a temporal resolution in our patient study group of between 85 and 155 msec, which provided a well-defined slice sensitivity profile and yielded 0.6-mm in-plane reconstructions.
|
|
Functional Analysis
Short-axis left ventricular data sets were evaluated semiautomatically
three times using a commercially available software application (Cardiac
Review, Brilliance Workspace, version 2.0 B2, Philips Medical Systems) with an
interval of more than 1 week between each measurement repetition by two
radiologists: one was considered experienced, with more than 5 years of
experience in cross-sectional cardiac imaging, and the other was considered
less inexperienced, with 1 year of experience in cross-sectional imaging.
Phantom studies validating the accuracy of volumetric segmentation of the used
software application in correlation with underlying heart rate were performed
before this study [12].
Ventricular endocardium and pericardium were assigned manually by selecting a
"seed point" in the center of the ventricular cavity and drawing a
"seeding oval" in the center of the myocardial wall, respectively.
Subsequently, the reviewers manually determined the most basal ventricular
image defined as the section in which the left ventricular myocardium extended
over at least 50% of the circumference of the left ventricular chamber
[5]. Finally, automated
segmentation techniques detected without further manipulations and corrections
of endocardial-pericardial interfaces after low-contrast thresholds were
defined [9]. Cases of evident
segmentation failures were noted.
Myocardial segments were localized with reference to both the long axis of the ventricle and the circumferential locations on the short-axis data sets in the basal, mid cavity, and apical even-parity regions of the left ventricle (Fig. 1A). The left ventricular segmentation resulted for each of the basal, mid-cavity, and apical regions in six segments—that is, the anterior segment; two septal segments, anteroseptal and inferoseptal; and three posterior wall segments, inferior, inferolateral, and anterolateral, for a total of 18 segments per ventricle [13] (Fig. 1B).
The software applications calculated for each segment the ventricular wall thickening expressed as a percentage of the volume increase of the individual wall segment in relation to the end-diastolic wall thickness. Furthermore, ventricular wall motion was assessed by evaluating the shift parallel to the long-axis centerline of each wall segment expressed in millimeters [14]. The durations of the postprocessing procedures for the entire data set were noted for each reviewer.
Statistical Analysis
Interobserver reliability was determined using the Bartko-Carpenter
approach, calculating the interobserver coefficient kappa for wall
thickening and wall motion of every cardiac segment
[15].
To assess the reproducibility of ventricular wall thickening and wall motion measurement results in repeated evaluation series to test the method of analysis, the coefficients of variance were calculated for both parameters, wall motion and wall thickening, in every ventricular segment for the entire study group. The coefficients of variance expressed the SD as a percentage of the arithmetic mean.
Subsequently, the paired Student's t test was used to explore whether any detected bias was related to the underlying methods of visualization, either CT or MRI.
Finally, the magnitude of the coefficients of variance were correlated to the underlying heart rate. This correlation was performed by creating scatterplots showing heart rate versus the magnitude of the coefficients of variance. If the plotted data were not linearly related, different curve estimation regression models, such as quadratic, cubic, logarithmic, or exponential models, were used to identify the specific type of regression and the regression parameters, such as the regression coefficient and significance level.
After measurement reproducibility was confirmed, measurement accuracy was evaluated by calculating the absolute measurement differences of the data obtained from either CT or MRI. The paired Student's t test was used to determine whether any detected bias was significantly related to the underlying methods of visualization. The paired Student's t test was also used to evaluate the entire patient study group, including all heart frequencies and a subset of patients with a heart rate of greater 100 bpm.
Thereafter, the magnitude of the absolute differences was correlated to the underlying heart rate. Scatterplots were analogously created to plot heart rate versus the magnitude of the absolute measurement difference for every ventricular segment. In cases of data points dispersed nonlinearly, curve estimation regression identified the specific type of model with its characteristic parameters, regression coefficient, and significance level.
All statistical analysis was performed using SPSS software (version 11.5, Statistical Package for the Social Sciences); a significance level of 0.005 was considered to be statistically significant.
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
Reproducibility of Wall Motility Analysis
Coefficients of variance identifying measurement variability on repeated
assessments of ventricular wall thickening and motion presented a dependency
of measurement variability that was related to segmental ventricular
localization. These detected measurement variations were observed uniformly in
CT and MR data sets (Table
1).
|
Measurement variability in septal (anteroseptal and inferoseptal) and anterior ventricular segments of CT and MR data sets averaged 6.6% ± 3.8% and ranged from 2.6% to 16.6% when evaluating ventricular wall thickening and presented an average of 10.4% ± 5.7%, with a range of 5.0-26.1%, when evaluating ventricular wall motion. On the other hand, measurement variability determined in the posterior wall segments (inferior, inferolateral, and anterolateral) averaged 4.1% ± 1.3% and ranged from 2.4% to 6.7% when evaluating ventricular wall thickening. Analogously, the average measurement variability in ventricular wall motion analysis averaged 6.3% ± 1.8%, ranging from 3.4% to 8.5% (Figs. 2A and 2B).
|
|
When comparing CT and MR data sets of left ventricular wall thickening by means of the paired Student's t test, neither a statistical assessment of circular regions at the base, mid cavity, or apex of the ventricle nor a statistical evaluation of the segmental localization (anterior, anteroseptal, inferoseptal, inferior, inferolateral, and anterolateral) presented statistically significant measurement variations between the two imaging techniques (p > 0.005). Similarly, a comparative assessment of the CT and MR data sets focusing on left ventricular wall motility was not able to identify statistically significant measurement variations between the two imaging techniques—neither in the circular ventricular regions nor in the left ventricular segments (p > 0.005).
Finally, a correlation of the magnitude of coefficient of variance with underlying heart rate resulted in quadratic regression coefficients ranging from 0.001 to 0.577 and from 0.007 to 0.652 for the CT and MR data sets, respectively, without statistical significance (p > 0.05) when evaluating left ventricular wall thickening. When focusing on left ventricular wall motility, quadratic regression coefficients ranging from 0.005 to 0.556 and from 0.03 to 0.585 for the CT and MR data sets, respectively, and significance levels (p values) of greater than 0.05 were obtained.
Therefore, the analysis of measurement reproducibility of left ventricular wall motility proved that data variability was dependent only on anatomic localization of the analyzed ventricular segment, not on either the imaging technique or on the underlying heart rate.
Accuracy of Wall Motility Analysis Depending on Imaging Technique
The absolute measurement differences of CT and MR data sets focusing on
left ventricular wall thickening (a percentage) and wall motion (in
millimeters) showed no segmental-dependent variation
(Table 2).
|
The direct comparison of CT and MR data sets for evaluating left ventricular wall thickening resulted in an average difference of 4.2% ± 1.5%, ranging from 1.2% to 6.9%. Comparing the CT and MR data sets for assessing left ventricular wall motion, an average difference of 0.8 ± 0.3 mm, ranging from 0.4 to 1.6 mm, was obtained.
A correlation of the absolute measurement differences for assessing left ventricular wall thickening with the underlying heart rate resulted in a quadratic regression coefficient of 0.971 and significance levels (p) of less than 0.0001 (Fig. 3A). Correlating the absolute measurement differences of CT and MRI for evaluating left ventricular wall motion, a quadratic regression coefficient of 0.975 and significance levels (p) equal to 0.0001 were calculated (Fig. 3B).
|
|
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
Functional cardiac imaging relies on ECG-gated image data sets obtained throughout the cardiac cycle, thereby emphasizing the need of verification, if CT and MRI can be used synergistically. For this study, we investigated the impact of using two imaging techniques to determine ventricular motility parameters such as ventricular wall thickening and motion, while correlating the observed measurement variability with the underlying heart rate. In a cardiac cine MRI study performed in a group of healthy volunteers, Miller et al. [29] called for a spatial resolution of less than 3-mm slice thickness and temporal resolution of less than 50 msec in ventricular motility analysis for accurate volumetric results. Advances in CT detector technology, gantry rotation times and 3D, and half-scan (180°) reconstruction algorithms were able to increase spatial resolution significantly while achieving moderate increases in temporal resolution, thus remaining more sensitive to cardiac motion compared with fast MRI [30]. By involving patients with various cardiac diseases, adding CT as a secondary imaging technique, and differentiating the ventricular myocardial walls into various segments, we evaluated the resulting effects, their localization, and their significance on ventricular motility parameters due to less than optimal temporal resolution in CT called for in prior studies, but with a significant increase in spatial resolution in MRI compared with standard cardiac MRI protocols.
This study was able to prove that an assessment of ventricular motility throughout the entire cardiac cycle and additional automated ventricular segmentation without further manipulations and corrections by the reviewers identified certain anatomic locations where significant intraobserver variations occurred. In particular, the anterior and septal ventricular segments, characterized by a thinner wall thickness compared with the posterior walls, proved to be susceptible to intraobserver measurement variability. Whereas the automatic endocardial segmentation homogeneously identified interfaces with equal-density threshold values in measurement, the prior determination of these threshold values was based on the manually drawn seeding oval. Therefore, during measurement repetitions with newly defined seeding ovals, slight variations of threshold values led to exponential and significant error propagation during 3D volumetric assumptions. The need for multiple measurement repetitions and subsequent data averaging was thereby emphasized. On the other hand, the thinnest physiologic ventricular wall segment, the anteroseptal segment representing the aortic outflow tract, might be considered an efficient segment for estimating measurement reproducibility while performing ventricular motility analysis.
By focusing on the comparison of the image data sets obtained with MRI and CT, we found that in general no statistically significant differences were identified when ventricular motility was evaluated on the basis of the data sets of either imaging technique in normofrequent patients. The lack of required temporal resolution in CT seemed to have been partially compensated by a significant increase in spatial resolution in ECG-gated data sets of normofrequent patients.
Impaired reproducibility in delineating ventricular interfaces in anterior and septal ventricular segments during repeated series was observed in data sets of both imaging techniques. However, the underlying reasons for this segmentation mismatch were of different origin: the MR data sets presented with variations in segmented pericardial and myocardial borders, whereas the CT data sets showed segmentation variations only along the myocardial interfaces.
Absolute segmental variations of ventricular motility parameters were found in motility assessments based on MR and CT data sets. Whereas the reasons of image degradation leading to segmentation mismatches might be of multiple origins—such as lack of required temporal or spatial resolution, different methods of contrast development combined with specific and varying sensitivities to cardiac and breathing motion and to susceptibility artifacts induced by the surrounding lung tissue in MRI—no statistically significant differences in ventricular motility parameters were found throughout the patient study group including a wide bandwidth of heart frequencies. Focusing however on the findings for patients with tachycardia, our results proved statistically significant data discrepancies between MR-based and CT-based wall motion assessments are due to the insufficient temporal resolution that is available with the current CT technology.
This study therefore proved the feasibility of the synergistic use of cardiac MR and CT data sets, which might have been originally and intentionally acquired to confirm other specific cardiac diagnoses, for ventricular motility analysis in normofrequent patients, therefore avoiding the administration of nephrotoxic contrast material and application of a radiation dose significantly higher compared with conventional cardiac angiography [31]. However, new dose modulation techniques, such as ECG-pulsed tube current modulation, have already proved to significantly reduce radiation dose in cardiac CT [32]. Ventricular motility analysis characterized by parameters, such as ventricular thickening and motion, thereby represents an assessment, which can be initially obtained and followed up with either of the imaging techniques evaluated in this study of normofrequent patients.
However, the entire patient group examined in this study was characterized by normorhythmic patterns. To emphasize the impact of physiologic factors such as heart frequency on cardiac imaging, this study was able to prove the existence of a range of optimal heart rates for the acquisition of CT image data sets. A close analysis of these image data sets showed that high cardiac frequencies—that is, those above approximately 100 bpm—were responsible for significant measurement variations, specifically a mismatch of wall thickening data of greater than 3% and a difference in segmental wall motion data of greater than 1 mm in a direct CT-MRI comparison. Even higher heart rates proved the existence of exponential error propagation, thus emphasizing the necessity of reporting ventricular motility results in combination with underlying heart frequency.
Our study has several limitations that must be addressed. Due to consecutive patient recruitment without exclusions, the patient study group presented a wide variety of cardiac diseases and conditions. This inhomogeneous study group did not allow a pathologically oriented analysis of the acquired CT and MR data sets. This heterogeneity in cardiac conditions in this study, on the other hand, emphasizes the robustness of recent imaging techniques and image postprocessing applications.
The relatively small number of patients in the entire study population showed physiologic frequency and rhythm patterns. More extreme variations in the frequency and rhythm patterns might lead to even greater data variability.
In conclusion, this study proved that cardiac motility assessment based on ECG-gated CT and MR data sets resulted in comparable ventricular function results for normofrequent patients; however, the high spatial resolution of cardiac MDCT cannot compensate for the lack of temporal resolution in tachycardic patients, thus emphasizing the necessity of reporting ventricular motility analysis results in combination with heart rate to allow consideration of this possible cause for measurement variation.
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
This article has been cited by other articles:
|
|
 |
|
  P. M. Colletti Cardiac Imaging 2006 Am. J. Roentgenol., June 1, 2006; 186(6_Supplement_2): S337 - S340. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
