| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
1 Department of Radiology, Division of Surgery, University of Vienna Medical
School, General Hospital Vienna, Waehringer Guertel 18-20, 1090 Vienna,
Austria.
2 Department of Radiology, Harborview Medical Center, 325 9th Ave., Box 359728,
Seattle, WA 98104.
3 Department of Radiology, Division of Osteology, University of Vienna Medical
School, General Hospital Vienna, 1090 Vienna, Austria.
Received August 23, 2002;
accepted after revision November 8, 2002.
Address correspondence to V. M. Metz.
Abstract
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
MATERIALS AND METHODS. A cadaveric head with artificial blunt facial trauma was examined using a four-channel MDCT scanner. The influence of acquisition parameters (collimation, 2 x 0.5 mm, 4 x 1 mm, 4 x 2.5 mm; tube current, 120 mAs, 90 mAs, 60 mAs), image reconstruction algorithms (standard vs ultra-high-resolution modes; reconstructed slice thicknesses, 0.5 mm, 1 mm, 3 mm; increment, 0.3 mm, 0.6 mm, 1.5 mm), and reformation algorithms (slice thicknesses, 0.5 mm, 1 mm, 3 mm; overlap, 0.5 mm, 1 mm, 3 mm) on detectability of facial fractures in multiplanar reformations with MDCT was analyzed.
RESULTS. Fracture detection was significantly higher with thin
multiplanar reformations (0.5 and 0.5 mm, 1 and 0.5 mm, and 1 and 1 mm)
(p 
0.014) acquired with 2 x 0.5 mm collimation (p
0.046) in ultra-high-resolution mode (p < 0.0005) with 120
mAs (p 0.025). Interobserver variability showed very good
agreement (
0.942). Non–ultra-high-resolution mode, lower
milliampere-seconds, and thick multiplanar reformations (3 and 0.5 mm, 3 and 1
mm, and 3 and 0.5 mm) showed significantly decreased fracture
detectability.
CONCLUSION. Although thin multiplanar reformations obtained from thin collimation (2 x 0.5 mm) are statistically superior for the detection of subtle fractures, 4 x 1 mm collimation is sufficient for routine diagnostic evaluation. Ultra-high-resolution mode with 120 mAs is mandatory for detection of clinically relevant fractures.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
MDCT allows the acquisition of very thin slices (e.g., 0.5 mm), resulting in high resolution not only in the axial plane (x- and y-axes) but also in the patient's longitudinal axis (z-axis). MDCT may approximate the so-called isotropic voxel, a cubic volume element that is a theoretic prerequisite for optimal two- and three-dimensional postprocessing of volume data sets [9]. Compared with state-of-the-art single-detector helical CT scanners, which also allow the acquisition of 0.5-mm-thick slices, the simultaneous use of more than one detector at least halves data acquisition time. In addition, a new interpolation algorithm, the Adaptive Axial Interpolation Algorithm [10], increases image quality. This versatile method works on parallel-beam data, generated by azimuthal rebinning, with helical interpolation performed by distance-dependent weighting. With this algorithm, slice-sensitivity profiles and pixel noise are constant for all pitch values in the given range by selection of appropriate weighting functions and suitable adjustment of the tube current. In addition, a broad number of reconstructed slice thicknesses can be generated from one given collimation [11]. Furthermore, the scanned volume can be increased without additional radiation dose penalty [12].
To our knowledge, scant literature addresses the potential roles for and limitations of MDCT in facial fractures. The aims of this study were to optimize acquisition protocols, multiplanar reformation algorithms, and radiation dose for the evaluation of facial fractures and to determine whether 2 x 0.5 mm collimation is necessary to depict the maximal number of fractures present. The influence of acquisition parameters (collimation, tube current), reconstruction (image reconstruction algorithms—including standard vs ultra-high-resolution modes, reconstructed slice thickness, increment), and reformation (slice thickness and overlap) on detectability of facial fractures in multiplanar reformations with MDCT was analyzed. In addition, the effects of algorithm and parameters on image noise, artifacts, and delineation of soft tissues were evaluated.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Facial fractures were artificially produced by means of blunt force. A steel block with rounded edges was placed parallel to the nose over the left orbit and struck once by a standard steel hammer. A second trauma site was produced similarly at the area of the left zygomatic arch. A total of 10 fractures with different amounts of dislocation, fragmentation, and extension into surrounding structures were created (Table 1).
|
CT was performed using a four-channel MDCT scanner (Somatom Plus 4 Volume Zoom, Siemens, Erlangen, Germany). The head was positioned at the isocenter of the CT scanner [13] and taped on the table to minimize motion artifacts due to table movement.
Acquisition Protocols
The term "collimation" describes the size of the detector
elements in the z-axis, the term "slice thickness" is the
full-width at half-maximum of the calculated slice sensitivity profile, and
the term "increment" describes the distance in millimeters from
the center of one slice to the center of the next slice; if the increment has
the same value as the slice thickness, no overlap occurs.
Examinations of the traumatized head were performed with three collimations (Table 2): First, with a 2 x 0.5 mm collimation with table feed of 1.3 mm, tube voltage of 120 kV, tube current of 120 mAs, and reconstructed slice thickness and reconstruction increment of 0.5 and 0.3 mm. Acquisition and reconstruction were performed in ultra-high-resolution mode. Second, we used 4 x 1 mm collimation with table feed of 4.5 mm; tube voltage of 120 kV; tube current of 120 mAs, 90 mAs, and 60 mAs and reconstructed slice-thickness and reconstruction increment of 1.0 and 0.6 mm. Acquisition and reconstruction were performed in ultra-high-resolution mode and non–ultra-high-resolution mode. Third, we used 4 x 2.5 mm collimation with table feed of 11.3 mm, tube voltage of 120 kV, tube current of 120 mAs, reconstructed slice-thickness and reconstruction increment of 3 and 1.5 mm. Acquisition and reconstruction were performed in non–ultra-high-resolution mode.
|
Rotation time was 0.75 sec in all protocols to obtain the maximal number of interpretations per rotation. The 4 x 1 mm collimation was used as a reference scan for comparison of different dosages, using 120 mAs, 90 mAs, and 60 mAs. The ultra-high-resolution mode was solely for high-resolution studies. For the non–ultra-high-resolution mode series, the H70 very sharp image reconstruction algorithm was used. For the 4 x 2.5 mm collimation, the scanner does not provide ultra-high-resolution mode.
With 2 x 0.5 mm collimation, the scan was divided into three parts because the scanning exceeded the maximal scanning time of 100 sec. These parts were acquired over contiguous volumes with the same spacing, field of view, and isocentral x and y coordinates. Because the head did not move, no misregistration was possible, and the three parts could be assembled into one volume data set according to table position.
Reformation Protocols
Of the axial volume data sets, coronal (n = 39) and sagittal
(n = 39) multiplanar reformations were generated with the following
parameters (Table 2): slice
thicknesses of 0.5 mm, 1 mm, 3 mm, and overlap of 0.5 mm, 1 mm, 3 mm. In
total, 78 reformatted series were generated on a Volume Wizard workstation
(Siemens). We used the following algorithms: thin multiplanar reformations
(0.5 and 0.5 mm, 1 and 0.5 mm, and 1 and 1 mm) generated by trilinear
interpolation without sub-sampling. For thick multiplanar reformations, (3 and
0.5 mm, 3 and 1 mm, and 3 and 3 mm), fusion of the thin multiplanar
reformations that were generated by trilinear interpolation produced images in
which the value of each pixel of the resulting thick multiplanar reformation
was the mean value of the corresponding pixels on the original thin
multiplanar reformations. Because of the sampling strategy used to create thin
multiplanar reformations, not all voxels on the rendering slab contributed to
the resulting reformatted image.
To isolate the independent influence of multiplanar reformations on the detection of subtle facial fractures, those fractures visible on all planes (n = 4) were excluded. Only those six fracture locations that were not visible in at least one plane (axial, n = 1; coronal, n = 2; sagittal, n = 3) were included in this study (Table 3).
|
Scoring System and Statistical Analysis
Fracture locations were scored by five experienced radiologists on the
original axial data set and the reformatted coronal and sagittal multiplanar
reformations. The reviewers scored the detectability of each of the six
fractures in 83 respective series (five axial, 39 coronal, and 39 sagittal)
with different parameters. All reviewers were aware of the fracture locations
in advance but unaware of the purpose of the study and of the acquisition and
reformation details of the respective series.
To asses fracture detection, noise, artifacts, and soft-tissue delineation, we used the following 3-point scoring system: fractures were scored with 0 for no fracture, 1 for suspicious fracture, and 2 for clear evidence of fracture. Noise and artifacts were scored with 0 for interfering noise and artifacts, 1 for disturbing noise and artifacts, and 2 for absence of disturbing noise and artifacts.
Delineation of soft tissues was scored with 0 for poor delineation of soft tissues, 1 for limited delineation of soft tissues, and 2 for excellent delineation of soft tissues. Evaluation of the reformatted series was performed on a Magic View workstation (Siemens). The dependent variable was the total fracture score on an interval scale. Independent variables were collimation, multiplanar reformation algorithm, tube current, and use of ultra-high-resolution mode.
We tested the following hypotheses: fracture detection accuracy increases
with thinner collimation, thinner multiplanar reformations, increasing tube
current, and use of ultra-high-resolution mode. In the first part, pooled data
analysis, the whole data set was tested for significant differences in
fracture score for each of the four independent variables. Because the whole
data set was tested four times, a Bonferroni adjustment was made to the level
of significance, reducing it to 
' equals 0.0125. In the second
part, individual series analysis, the differences in fracture scores of each
individual sequence were assessed in the same manner. In these calculations,
the level of significance alpha was 0.05 in all calculations. Different scores
of the respective series were tested with the Wilcoxon's signed rank test,
which is a nonparametric test for the significance of the difference between
the distributions of two nonindependent samples such as repeated measures or
matched pairs, on a standard personal computer with a statistics software
package (SPSS 10.0, Chicago, IL). Different scores of the respective observers
were expressed as interobserver variability, quantified with Cohen's Kappa
test [14]. The strength of
agreement ratings were classified as poor ( < 0.2), fair (
= 0.21–0.40), moderate ( = 0.41–0.60), good ( =
0.61–0.80), and very good ( = 0.81–1.00)
[15].
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
For interobserver variability, the kappa value was 0.942 (very good agreement) for the axial data sets, the kappa value was 0.947 (very good agreement) for the coronal reformatted data sets, and the kappa value was 0.953 (very good agreement) for the sagittal multiplanar reformations.
Detectability of fractures on the axial CT series acquired with 2 x 0.5 mm collimation was not significantly superior to the 4 x 1 mm collimation with 120 mAs in ultra-high-resolution mode (p = 0.157). But the 2 x 0.5 mm collimation was superior to the 4 x 1 mm collimation with 120 mAs in non–ultra-high-resolution mode (p = 0.046) and to the 4 x 2.5 mm collimation with 120 mAs (p = 0.038) in non–ultra-high-resolution mode.
Pooled Data Analysis
Multiplanar reformations obtained from series acquired in
ultra-high-resolution mode were significantly superior to those acquired in
non–ultra-high-resolution mode (p < 0.0005). Multiplanar
reformations obtained from series acquired with 120 mAs were significantly
superior to those acquired in the 90 mAs series (p < 0.0005) and
in the 60 mAs series (p < 0.0005). In addition, detectability of
fractures on the 90 mAs series was significantly higher compared with the 60
mAs series (p < 0.0005). Multiplanar reformations obtained from
series acquired with 2 x 0.5 mm collimation were significantly superior
to those acquired with 4 x 1 mm (p < 0.0005) and 4 x
2.5 mm collimation (p < 0.0005). The 4 x 1 mm collimation
was significantly superior compared with the 4 x 2.5 mm collimation
(p < 0.0005).
Very thin 0.5- and 0.5-mm multiplanar reformations were not significantly superior to 1- and 0.5-mm (p = 0.317), 1- and 1-mm (p = 0.180), 3- and 0.5-mm (p = 0.059), and 3- and 1-mm (p = 0.015) multiplanar reformations but were superior to 3- and 3-mm multiplanar reformations (p = 0.001); 1- and 0.5-mm multiplanar reformations were significantly superior to all other multiplanar reformations (p < 0.0005), except 0.5- and 0.5-mm reformations.
Individual Series Analysis
Comparing the different multiplanar reformation series, this study shows
that multiplanar reformations in ultra-high-resolution mode are generally
superior to corresponding series in non–ultra-high-resolution mode
(p < 0.0005) (Figs.
1A,
1B). Series with 120 mAs were
significantly superior to the series with 90 mAs (p 0.025) and
to the series with 60 mAs (p 0.001); series with 90 mAs were
significantly superior to those with 60 mAs (p 0.034) (Figs.
2A,
2B,
2C). Acquisition with 2 x
0.5 mm collimation was superior for multiplanar reformation purposes compared
with 4 x 1 mm and 4 x 2.5 mm collimation (p 0.046)
(Figs. 3A,
3B,
3C).
|
|
|
|
|
|
|
|
Concerning the reformation algorithms, 0.5- and 0.5-mm multiplanar
reformation was not significantly superior to 1- and 0.5-mm and 1- and 1-mm
reformation (p > 0.059). However, all three reformations (0.5 and
0.5 mm, 1 and 0.5 mm, and 1 and 1 mm) were significantly superior to all other
multiplanar reformation series (p 0.014) (Figs.
4A,
4B,
4C,
4D).
|
|
|
|
Additional Results
Noise and soft-tissue delineation did not influence overall scores of the
series. However, the handling of thin multiplanar reformations (e.g., 0.5 and
0.5 mm) on the reviewing workstation leaves much to be desired because of low
display-system speed with the high number of images.
Each multiplanar reformation in this study was available to the radiologist for review in 10–15 min after completion of the axial scans at the postprocessing workstation and in 40–50 min after completion of the axial scans at the reviewing workstation.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
However, multiplanar reformations from conventional helical CT often have artifacts because of low image resolution in the z-axis. Concerning multiplanar reformations, the general rule is that the smaller the distance between the slices and the larger the degree of overlap of the original images, the greater the resolution in the examination direction (z-axes). The in-plane resolution of MDCT (x- and y-axis) may contribute to the diagnostic utility of the reformatted series (z-axis) despite the lack of significant difference between the 2 x 0.5 mm and the 4 x 1 mm collimation in ultra-high-resolution mode as was found in the axial series.
To objectively determine the spatial resolution capabilities of a CT
system, we most commonly used the modular transfer function, which shows the
frequency components of a given structure in line pairs per centimeter. The
achievable spatial resolution for a given system is most commonly specified
with the frequency value at 2% of the modular transfer function. At a rotation
time of 0.75 sec, the in-plane resolution for the U90 ultrasharp image
reconstruction algorithm, manifests a modular transfer function at 2%
(
2) of 22.63 line pairs per centimeter and for the H70 very
sharp image reconstruction algorithm a 2 of 14.89 line pairs
per centimeter (Vestner H, Siemens, Forchheim, Germany, personal
communication). Currently the manufacturer of the scanner could not provide
similar data for the z-axis resolution for the image reconstruction
algorithms used in this study, but data were provided for two similar image
reconstruction algorithims. The sharp B70s image reconstruction algorithm
shows, at collimated slice thickness (dcoll) of 0.5 mm and a
reconstruction increment for the longitudinal axis (
Zinc) of
0.3 mm, an in-plane modular transfer function at 50%
(50xy) of 9.2 line pairs per centimeter, whereas the
longitudinal modular transfer function at 50% (50z)
is 8.2 line pairs per centimeter, which is 81% of
50xy. For the smooth B40s image reconstruction
algorithm (dcoll = 1.0 mm, Zinc = 0.6 mm),
50xy is 4.7 line pairs per centimeter, and
50z is 4.1 line pairs per centimeter, which is 87% of
50xy (Wallschlager H, Siemens, Erlangen, Germany,
personal communication).
Moreover, the quality of the resulting series increases with narrow slice collimation [9]. The results of this study show the superiority of small-increment multiplanar reformation protocols derived from thin-collimation axial-scan protocols in the detection of facial fractures using four-channel MDCT.
Subtle fractures of the facial bones are best shown on thin reformations (e.g., 0.5 and 0.5 mm) of thin acquisitions (2 x 0.5 mm) obtained with 120 mAs. For example, a dislocation of fragments of the fractured lateral maxillary sinus wall was clearly visible on the coronal 0.5- and 0.5-mm reformation but was not shown on the other reformation series. However, we found no clear incremental advantage of using 0.5- and 0.5-mm, 1- and 0.5-mm, or 1- and 1-mm reformations. This may be due to the decreased signal-to-noise ratio for the 0.5- and 0.5-mm reformation compared with that of the 1- and 0.5-mm and the 1- and 1-mm reformations. The signal-to-noise ratio was only subjectively evaluated by the reviewers.
Ultra-high-resolution mode improves delineation of thin osseous lamellae and increases significant fracture delineation by changing effective detector aperture (i.e., collimation) while using high-contrast, high-spatial-frequency image reconstruction algorithms. A combination of flying focal spot and quarter-detector shift effectively increases the in-plane sampling by a factor of four and allows considerably improved in-plane resolution, although the depiction of soft-tissue is restricted.
Our data show that accurate imaging of subtle fractures and fracture dislocations is not reliable with a tube current lower than 120 mAs. With optimal MDCT protocols, dosage can be reduced by up to 33% compared with conventional helical CT [17, 18]. In this study, the radiation dose was substantially lower than that used for typical musculoskeletal imaging applications with MDCT [19] and was also lower than the proposed settings from the manufacturer for 140 kV and 160 mAs (Table 2). High-quality multiplanar reformations obtained from thin axial data sets allow omission of direct coronal scans and result in a further substantial reduction of radiation dose.
A potential limitation of thin collimation (2 x 0.5 mm) is that concentration of subvolume acquisitions are required to achieve the desired volume coverage (total scanning time > 100 sec). In part, this problem derives from the experimental design in which scanning of the entire head was desired. In clinical practice, the number of resulting images (Table 4) would be reduced by limiting scanned volume to the face, which would facilitate image acquisition and handling and thereby improve cost-effectiveness. Eight- and 16-channel MDCT scanners make possible high-resolution scans of the entire skull and face in one acquisition if indicated.
|
Although some fractures generated in this study may be of minor or even no clinical relevance, our aim was to depict all the fractures to achieve reliable protocols for future examinations of patients. In practice, scanning of the face with 4 x 1 mm collimation gives enough reliable information for treatment. In case of inconclusive findings, 2 x 0.5 mm collimation may be added.
As a result of the design of this study, assessments of sensitivity cannot be generalized to clinical populations because all observers agreed on the presence of the described fracture locations (Table 1). Furthermore, no additional fractures besides those at the described fracture locations were found by the reviewing radiologists, and the absence of normal cases precludes determination of specificity.
Each multiplanar reformation in this study was available for review by the radiologist soon after completion of the axial scans. This availability supports the routine use of axial scans with multiplanar reformation even when immediate reports are needed. In the General Hospital Vienna, postprocessing of images (e.g., multiplanar reformation) is not performed on the control unit of the scanner but rather on a second independent workstation. This off-line work serves to maintain the primary service capacity for each CT scanner.
However, for efficient handling of many images, more sophisticated soft- and hardware tools are needed. Although the workstation used in this study allows at least 10 images per second in cine mode, the reviewing process of the thin ultra-high-resolution series is still time-consuming.
In conclusion, we do not need 2 x 0.5 mm collimation for routine diagnostic evaluation of facial fractures although 2 x 0.5 mm collimation showed statistical superiority over all other collimations, but all fractures additionally found with the 2 x 0.5 mm collimation compared with the 4 x 1 mm collimation were of no clinical relevance for this study.
For multiplanar reformation of volume data sets acquired with MDCT, thin reformations (0.5 and 0.5 mm, 1 and 0.5 mm, and 1 and 1 mm) in ultra-high-resolution mode are preferred and can minimize diagnostic errors for subtle fractures. Although there is no statistically significant difference between the 0.5 and 0.5 mm, 1 and 0.5 mm, and 1- and 1-mm reformation, there were fracture-displacements that could only be delineated with 0.5- and 0.5-mm reformation. In clinical practice, however, all relevant fractures are present on the 1- and 0.5-mm reformation. In addition, acquisition should be performed with at least 120 mAs.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
This article has been cited by other articles:
|
|
 |
|
  T. Miyoshi, M. Kanematsu, H. Kondo, S. Goshima, Y. Tsuge, A. Hatcho, Y. Shiratori, M. Onozuka, N. Moriyama, and K. T. Bae Abdomen: Angiography with 16-Detector CT--Comparison of Image Quality and Radiation Dose between Studies with 0.625-mm and those with 1.25-mm Collimation Radiology, October 1, 2008; 249(1): 142 - 150. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 P.M. Phal, L.P. Riccelli, P. Wang, G.M. Nesbit, and J.C. Anderson Fracture Detection in the Cervical Spine with Multidetector CT: 1-mm versus 3-mm Axial Images AJNR Am. J. Neuroradiol., September 1, 2008; 29(8): 1446 - 1449. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 S E J Connor and N Chaudhary Imaging of maxillofacial and skull base trauma Imaging, March 1, 2007; 19(1): 71 - 82. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 T Rodt, S. Bartling, J. Zajaczek, M. Vafa, T Kapapa, O Majdani, J. Krauss, M Zumkeller, H Matthies, H Becker, et al. Evaluation of surface and volume rendering in 3D-CT of facial fractures. Dentomaxillofac. Radiol., July 1, 2006; 35(4): 227 - 231. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 C. Kulinna, R. Eibel, W. Matzek, H. Bonel, D. Aust, T. Strauss, M. Reiser, and J. Scheidler Staging of Rectal Cancer: Diagnostic Potential of Multiplanar Reconstructions with MDCT Am. J. Roentgenol., August 1, 2004; 183(2): 421 - 427. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
