| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
1 Department of Radiology, University of Vienna, Waehringer Guertel 18-20,
Vienna A-1090, Austria.
2 Department of Surgery, University of Vienna, Vienna, Austria.
3 Department of Biomedical Engineering and Physics, University of Vienna,
Vienna, Austria.
4 Department of Radiology, Massachusetts General Hospital and Harvard Medical
School, 32 Fruit St., Boston, MA 02114.
Received July 2, 2003;
accepted after revision September 23, 2003.
Address correspondence to W. Schima
(wolfgang.schima{at}univie.ac.at).
Abstract
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
MATERIALS AND METHODS. Surgically resected adrenal tumor specimens (total, seven; adenomas, three; nonadenomas, four; size range, 17–76 mm), were placed in an anthropomorphic phantom. Lesion specimens were scanned with one MDCT and two single-detector scanners. Scanning protocols for all three scanners included variations in kilovoltage (140, 120, and 80 [Somatom Plus 4, Somatom VolumeZoom] or 100 [Tomoscan AV] kVp) and slice thickness. Hounsfield measurements were performed on exactly matched slices using regions of interest of a constant size.
RESULTS. The difference in lesion Hounsfield measurements among scanning protocols with 140, 120, and 100/80 kVp was up to 6.2 H for the adenoma group and up to 3.8 H for the nonadenoma group. The comparison of the Tomoscan AV and the Somatom Plus 4 scanners showed a mean difference of 2.6 H at 120 kVp and of 4.6 H at 140 kVp. The differences between the Tomoscan AV and Somatom VolumeZoom scanners were 1.7 and 3.6 H for 120 and 140 kVp, respectively. Between the two Somatom scanners, the divergence was 2.9 and 3.3 H for the two kilovoltage settings. Differentiation between adenomas and nonadenomas was better at lower kilovoltage. Slice thickness did not affect the CT density measurements significantly.
CONCLUSION. Significant differences in CT density measurements of adrenal tumors may occur when different CT scanners or imaging protocols are used. The dependence of measurements on scanner type and scanning technique makes the recommendation of a universal, scanner- and protocol-independent threshold problematic.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
In this study we assessed the variability of CT attenuation values for adrenal tumors using a range of different scanning protocols and different CT scanners. Our purpose was to determine whether the establishment of a scanner-independent threshold for differentiation between adenomas and nonadenomas was feasible.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
We divided the lesions into two groups, the adenoma group and the nonadenoma group; the hemorrhagic adenoma was classified in the nonadenoma group.
Phantom and CT Technique
After excision, the surgical specimens were immediately and in toto welded
onto a polyethylene terephthalate foil. The specimens were embedded in a
water-filled anthropomorphic acrylic (polymethylmethacrylate) phantom (axial
diameter, 30 x 22.5 cm; length, 16.5 cm) and fixed on a plate having
attenuation equivalent to that of water, with a cadaver lumbar vertebra placed
in the center. The maximum transverse diameter of the fixed specimen ranged
from 1.7 to 7.6 cm.
All seven specimens were scanned with Tomoscan AV (Philips, Best, The Netherlands) and Somatom Plus 4 (Siemens, Erlangen, Germany) single-detector scanners. In four of the seven specimens, additional scans with a Somatom Volume-Zoom (VZ) multidetector scanner (Siemens) were obtained. CT scanning protocols for the three different scanners are summarized in Table 1.
|
The order of the scans was chosen randomly depending on the clinical availability of the scanners. In our study, the routinely performed calibration procedure was not altered for any scanners; therefore, air calibration was performed every 3 hr. The overall time of the three CT scans did not exceed 45 min.
Image Analysis
Attenuation measurements were performed on a commercially available PC
workstation using an Impax PACS (picture archiving and communication system)
(Agfa-Gevaert, Mortsel, Belgium). The attenuation was measured with circular
region-of-interest (ROI) cursors that were placed over the lesion
(Fig. 1). The ROI circle was
made as large as possible, excluding lesion edges to avoid partial volume
effects. The ROI measurements were obtained in three slices of the lesions,
with a gap of at least one slice between the three slices, to cover a larger
part of the lesion. The ROIs were placed in identical slices for all scanners,
and ROI size and position were constant for the compared slices. The mean and
SD of the lesion attenuation in Hounsfield units were measured. The
measurements of the three slices were averaged. In addition, the mean and SD
of water inside the phantom were obtained.
|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
|
|
Changing the slice thickness to 3 mm at 140 kVp changed the mean density levels 1.2 H or less. The mean densities of the water measurements in the phantom were –1.2, 0.4, and –1.8 H (at 100, 120, and 140 kVp, respectively), for the Tomoscan AV scanner; –0.2, 0.4, and 0.4 H (at 80, 120, and 140 kVp) for the Somatom Plus 4 scanner; and –2.4, –2.5, and –2.6 H (at 80, 120, and 140 kVp) for the Somatom VZ scanner.
The SD of the density measurements of the lesions differed slightly among the scanners but increased with a decrease in tube voltage (peak kilovoltage): from 9.4–14.4 H (at 140 kVp) to 12.4–37.5 H (at 80 kVp).
Interscanner Comparison
The comparison of the Tomoscan AV and the Somatom Plus 4 scanner at 120 kVp
showed a mean difference of 2.6 H (range, 0–5.7 H); at 140 kVp, the mean
difference was 4.6 H (range, 1.7–7.0 H). The mean differences between
the Tomoscan AV and the Somatom VZ scanners were 1.7 H (range, 0.3–3.3
H) and 3.6 H (range, 0.3–8.2 H) for 120 and 140 kVp, respectively.
Finally, the mean divergence between both Somatom scanners was 2.9 H (range,
0–4.3 H) and 3.3 H (range, 1.3–5.7 H) for both kilovoltage
settings. Table 3 summarizes
the interscanner comparison; the results are given by adenoma and nonadenoma
groups.
|
Intrascanner Comparison
For intrascanner comparison, the individual measurements of each lesion
obtained on the three different scanners were compared. The intrascanner
differences are listed in Table
4, which summarizes the mean differences of the individual lesions
according to kilovoltage variations for the three scanners. Again, the results
are given by adenoma and nonadenoma groups.
|
Combined Variation of Scanner Type and Kilovoltage
After variation of both scanner type and protocol, we observed a maximum
variation of the comparable mean group densities of 6.3 H (adenomas) and 8.4 H
(nonadenomas) (Table 2). The
maximum intraindividual variation of CT attenuation was 12 H.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
The use of a universal, scanner-independent threshold inherently assumes that different scanner types and different scanning protocols lead to constant density measurements [12]. Although some authors indicate the possibility of using absolute CT attenuation values for the characterization of tissue types [19, 20], some authors found, as we did, a wide range (>10 H) of CT attenuation for a particular tissue type, depending on the tube voltage (peak kilovoltage), scanner type, reconstruction algorithm, slice thickness, and tissue surrounding the lesion [21–24].
Three scanner vendors (General Electric Medical Systems [2, 3, 5–8, 10], Philips [4], and Siemens [1, 9]) were involved in most of the studies regarding differentiation of adrenal lesions. Details of scanning protocols were provided in only some of the studies; the kilovoltage values, in particular, were given by only a few authors [3, 9], with the value usually set at 120 kVp. Thus, none of the studies provided data about the influence of different techniques on density measurements.
In the literature, the slice thickness for adrenal CT scanning protocols ranges from 1 mm [8] to 4 [1], 5 [2, 4–10], and 10 mm [2–5, 7, 8]. If the slice thickness chosen is too thick in relation to the size of the lesions, then partial volume averaging may occur during ROI measurements [21–23]. In clinical practice, the slice thickness should be less than half the size of the lesion to avoid inaccuracies of measurement [22]. Therefore, a slice thickness of 3–5 mm seems appropriate for assessment of most adrenal masses [17]. In the present study, changing the slice thickness from 5 to 3 mm led to differences in CT densities of up to 1.2 H, using the same scanner and otherwise identical protocols. The most likely reason is that the inherent inhomogeneity of tumor tissue may lead to variations in Hounsfield density measurements if different slice thicknesses result in different tumor volume coverage in the z-axis.
Our study shows a protocol-dependent variation (different kilovoltage) in mean densities of up to 6.4 H (Table 2). In the nonadenoma group, Hounsfield values tended to decrease with rising kilovoltage. In the adenoma group, we observed an inconsistent kilovoltage influence, with increasing and decreasing values. The differences in Hounsfield units between the groups show an inverse proportional relationship to the kilovoltage for all scanners; that is, the differences between the groups are significantly smaller for higher kilovoltage values (Fig. 2). The energy dependence of CT attenuation values has been well described [22, 25, 26]. This fact can be explained by the different kilovoltage dependence of the attenuation coefficients, which is the result of the varying importance of the photoelectric and Compton effects. In general, the attenuation coefficients converge at high energies [27], which means that both absolute CT attenuation values and differences in CT attenuation values for identical lesions are kilovoltage-dependent. However, the reconstruction and filtering algorithms tend to hold the Hounsfield values constant, which can be achieved only for a certain atomic number, leading to a more complex relationship between Hounsfield measurements and kilovoltage.
Our results are in good accord with these theoretic considerations. In clinical practice, the magnitude of the attenuation difference between adenomas and nonadenomas has a direct influence on the sensitivity and specificity for the threshold suggested for differentiation. Because the attenuation coefficients (for low-density adenomas and high-density nonadenomas) converge at high kilovoltage values, our results suggest that the use of low kilovoltage values may increase the diagnostic value of a fixed density threshold.
The scanner-dependent variation of the mean densities for all lesions of one group, using fixed scanning protocols, showed a difference of up to 6.4 H (Table 2). The individual comparison of the lesions showed mean differences of up to 6.5 H (Table 3). The differences of the individual measurements were in most cases higher for the nonadenoma group than for the adenoma group, and higher at 140 kVp than at 120 kVp.
After varying both scanner and protocol, we observed a maximum intraindividual variation in CT attenuation of 12 H. This finding may have a serious impact on the correct classification of lesions with CT attenuation around the suggested threshold level of 10 H [12].
The absolute density measurements of lesions may also depend on the surrounding tissue [21, 22, 24]. Reconstruction algorithm [22, 23] and patient diameter [24] may have additional negative effects on the reproducibility of the CT attenuation, although these influences were not evaluated in our study.
A limitation of our study was the use of an anthropomorphic phantom and surgical specimens rather than in vivo examinations. However, radiation exposure for patients would not have been acceptable in this study design. In addition, the comparability of the measurements using different scanners is superior using a phantom with a fixed tissue specimen. Despite the fact that the specimens contained a smaller amount of blood than that present in vivo, the mean densities of both the adenomas and the nonadenomas were in the expected ranges. Assuming that the differences between the in vivo and ex vivo measurements are comparable, measurements of relative density differences are feasible. Although, as in clinical practice, scanner calibration was not performed immediately before the measurements, the mean deviation of the water measurements from the ideal value of 0 H was minimal, which confirms the quality of calibration procedures. The limited number of specimens reflects the difficulty of performing scanning of freshly resected surgical specimens on three scanners in a limited time.
In conclusion, our study confirms the benefit of using attenuation values of unenhanced CT for differentiation of adrenal lesions. Nevertheless, the dependence of measurements on scanner type and scanning technique makes problematic the recommendation of a universal scanner- and protocol-independent threshold.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
This article has been cited by other articles:
|
|
 |
|
  P. F. Hahn, M. A. Blake, and G. W. L. Boland Adrenal Lesions: Attenuation Measurement Differences between CT Scanners Radiology, August 1, 2006; 240(2): 458 - 463. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
= Somatom VolumeZoom scanner
(Siemens), 
= Tomoscan AV scanner (Philips, Best, The Netherlands).