| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Original Research |
1 Department of Radiology, Beth Israel Deaconess Medical Center/Harvard Medical
School, One Deaconess Rd., WCC-302, Boston, MA 02215.
2 PERFUSE Core Laboratory and Data Coordinating Center, Harvard Medical School,
Boston, MA.
3 Small Animal Imaging Facility, Harvard Medical School, Boston, MA.
Received October 23, 2006;
accepted after revision October 12, 2007.
Address correspondence to M. E. Clouse
(mclouse{at}bidmc.harvard.edu).
Abstract
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
MATERIALS AND METHODS. Polyethylene phantoms representing noncalcified plaque were scanned in both MDCT and micro-CT scanners and inter- and intrareader variability of volume calculation was performed.
RESULTS. Volume measurements by both MDCT and micro-CT were comparable to the true volume as measured by micrometry (< 3%, p = 0.05).
CONCLUSION. There appears to be no significant difference (< 3%) between MDCT and micro-CT measurements.
Keywords: aorta • CT angiography • coronary imaging • MDCT • micro-CT • voxel analysis
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
To compare the micro-CT measurements with those of clinical scanner studies, the same tubing containing the polyethylene particles was scanned in a standard 16-cm Lucite Head Phantom (Model 76–414, Cardinal Health) using an MDCT Aquilion 64 scanner (Toshiba America Medical Systems) with the following parameters: focal spot size, 1.4 x 1.6 mm; slice thickness, 0.5 mm; overlap, 0.2-mm; gantry rotation, 400 milliseconds; 135 peak kVp; 350 mA; field of view, 32 cm; image matrix, 512 x 512; pixel size, 0.39 mm2; continuous helical scanning; pitch, 0.2–0.3; without ECG input. Cross-sectional images were reconstructed using 400-µ voxels on the Analyze 6 workstation. Volume measurements were evaluated by three independent readers.
Volume Measurements
The reconstructed images from both micro-CT and MDCT were sent to a
stand-alone image processing workstation for volume measurement using our
voxel analysis technique: Step 1, representative cross-sectional images were
viewed by the readers and one image was selected containing all three regions
(particle, lumen, and wall). Step 2, Hounsfield density profiles were manually
plotted to encompass only the wall and lumen. Step 3, Hounsfield unit values
at these two boundaries (H1, inner wall and lumen; H2, outer wall and
epicardial fat) were determined by the computer on the basis of the gradient
changes. This step was repeated four times to obtain an average. Step 4, a 3D
intraluminal volume model was created based on the two Hounsfield values
obtained in Step 3. This model represents the sum of these pixels that exceed
the defined Hounsfield unit limits. Step 5, the wall of the 3D model was
peeled away, leaving only the Hounsfield density (volume) of the particle
(Figs. 1A,
1B,
1C,
1D,
2A,
2B,
2C, and
2D).
|
|
|
|
|
|
|
|
The true volumes of the intraluminal polyethylene particles were 9.50, 4.07, and 3.71 mm3 as measured by the micrometer. The accuracy of each measurement method was calculated as the aver age percentage of difference between measured volume and true volume, and the precision was similarly calculated as the average percentage of difference between the first and second volume measurements performed by the same person on the same particle by the same method. All percentages are calculated relative to the true volume of the particle. Accuracy and precision estimates for each scanning method are summarized per particle and across all polyethylene particles as mean and 95% CIs.
For each measurement method, the correlation of true particle size with either measurement accuracy or measurement precision was determined using a rank sum Kruskal-Wallis test for association. Tests for differences of accuracy or precision between methods were evaluated using Wilcoxon's rank sum test for association and the Student's t test for the difference between two means. The intraand interreader variability results are reported as the within-reader and between-reader SDs. The overall reliability of each measurement method is summarized as the concordance correlation coefficient [9, 10], which incorporates both within- and between-reader reliabilities. All analyses were conducted using Stata version 9.2 software (Stata Statistical Software).
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
|
The accuracy and precision errors of polyethylene particle volume measurements are summarized in Table 2. Overall volume measurements by each method (micro-CT and MDCT) were 2.05% less than the true volume after averaging over all the three particle sizes. The difference in accuracy between the two scanners was 0.00% (95% CI, –5.3% to 5.3%), with a p value (p = 0.25) for comparing the accuracy of micro-CT with that of MDCT. The magnitude of accuracy errors (as a percentage of true particle volume) did vary according to the particle volume for micro-CT (p = 0.03), whereas that with MDCT showed a weak trend (p = 0.14) (Table 2).
|
The precision error of volume measurements with micro-CT was 0.88% and with MDCT was 2.17%, so that overall precision error with MDCT was 1.29% greater than with micro-CT (95% CI, –4.0% to 6.6%). There was a possible trend for different precision errors over the range of true volumes when measured by micro-CT (p = 0.19), but there was otherwise little indication of any association between precision error and either the measurement method or the true volume (Table 2).
The intra- and interreader variabilities are summarized in Table 3. Intrareader SD was 0.098 mm [3] (range, 0.061–0.155 mm3) for micro-CT and 0.190 mm3 (0.119–0.301 mm3) for MDCT. Interreader SD was 0.038 mm3 (0.002–0.655 mm3) for micro-CT and 0.67 mm3 (0.377–1.223 mm3) for MDCT. The concordance correlation coefficient between intra- and interreader reliability was 0.999 (range, 0.997–1,000) for micro-CT and 0.972 (range, 0.948–0.995) for MDCT (Table 3).
|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Excellent concordance of volume measurements was seen in terms of intra- and interreader reliability for both micro-CT (concordance correlation coefficient = 0.999) and MDCT (concordance correlation coefficient = 0.972) (Table 3). In addition, the micro-CT and MDCT measurements were accurate relative to true volume measurements. Overall accuracy error was 2.05% (range, –10.44% to 4.00%) less than true volume for the three averaged polyethylene particles for both scanners. Although the individual accuracy errors varied, we found no significant difference between micro-CT (p = 0.025) and MDCT (p = 0.135). This is not unexpected because measurement using larger voxels could cause larger variations when measuring smaller particles due to volume averaging; thus, an individual voxel accounts for a larger percentage of the volume of a smaller particle. It may indeed be more accurate to measure all polyethylene particles with 100-µ rather than 400-µ voxels with the clinical scanner. For example, for the 4.07 mm3 polyethylene phantom particle, one voxel accounted for a volume error of 8.89% (Table 1).
Volume measurements by both micro-CT and MDCT were similarly precise (0.88% vs 2.17%), with a difference in precision error of 1.29% (95% CI, –4.0% to 6.6%). A possible trend was seen for difference in precision estimates over the range in particle size for micro-CT (p = 0.19), but there was otherwise little indication of any associated difference between precision, scanners, or true particle volume (p = 0.505) (Table 2). These findings, however promising, must be further evaluated with different voxel sizes for both scanners.
Our voxel analysis technique using Analyze 6 software for measuring noncalcified plaque has been shown by micrometry and micro-CT to be accurate. Apparently no significant difference exists in terms of volume measurements between micro-CT and MDCT using our technique. Both scanners appear able to image the particle sizes with excellent agreements: accuracy error, –2.05% (micro-CT and MDCT) and precision error, 0.88% (micro-CT) versus 2.17% (MDCT). Overall precision error between the two scanners was 1.29%. Similar to the volume measurement findings given earlier, the process requires further substantiation with a larger sample size.
Accuracy error for true volume (2.05% greater or less than the measured number) appears accurate and appropriate in the clinical setting. MDCT did underestimate true volume by 2.05% but, more important, MDCT was not different from micro-CT in the sample (mean, 0.0%). Thus, if the accuracy of the two methods were different, it is not likely to be significant (95% CI, –5.3% to 5.3%).
The precision error is also important. The average precision error for MDCT was 2.2%, which was not very different from that for micro-CT (mean difference, 1.3%; 95% CI, –4.0% to 6.7%). We doubt that a precision error of this amount would be significant in the context of measuring plaque in the clinical setting.
In addition to the previously mentioned accuracy in volume validation, another significant advantage of our voxel analysis technique is that it uses Hounsfield unit gradient changes, rather than fixed Hounsfield unit values, to determine the boundaries of the inner wall and lumen and of the outer wall and epicardial fat, which allows this technique to be patient-specific (i.e., adaptable to patient size).
On the basis of our phantom experiments, the results suggest that our technique is adequate for measuring the volume of noncalcified plaque in coronary arteries. These findings have allowed us to perform a pilot study measuring plaque in a small number of coronary arteries using 100-µ voxels in an institutional review board–approved study. The weakness of the study is that it was performed using a phantom model rather than histopathologic specimens, necessitating the use of air as contrast material outside and inside the tubes. In addition, the use of 100-µ voxels in clinical studies may improve the accuracy for volume and precision measurements.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
This article has been cited by other articles:
|
|
 |
|
  H. Brodoefel, C. Burgstahler, A. Sabir, C.-S. Yam, F. Khosa, C. D. Claussen, and M. E. Clouse Coronary Plaque Quantification by Voxel Analysis: Dual-Source MDCT Angiography Versus Intravascular Sonography Am. J. Roentgenol., March 1, 2009; 192(3): W84 - W89. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
