| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Original Research |
1 Institute of Diagnostic, Interventional, and Pediatric Radiology, University
of Berne Inselspital, Freiburgstrasse, Berne CH-3010, Switzerland.
2 Institute of Clinical Radiology, Ludwig-Maximilians-University Munich, Munich,
Germany.
3 Department of Gynecology and Obstetrics, University of Berne Inselspital,
Berne, Switzerland.
4 Department of Biostatistics, Tulane University, New Orleans, LA.
5 Siemens Medical Solutions, Forchheim, Germany.
Received July 22, 2004;
accepted after revision October 19, 2004.
Address correspondence to H. M. Bonel
(bonel{at}gmx.ch).
Abstract
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
MATERIALS AND METHODS. Thirty anatomic specimens were examined with an MDCT collimation of 4.0 x 1.0 mm and 2.0 x 0.5 mm (80, 120, and 140 kV; 80, 100, 130, 160, and 200 mA; rotation time, 0.5 0.75, 1.0 sec; pitch, 1.0, 1.3, 1.5, and 2.0). Coronal images were reconstructed using a slice thickness of 0.5, 1.0, and 2.0 mm with 60% overlap. Three observers evaluated all images independently for gross and fine anatomic detail. Diagnostic confidence was tested using Shrout-Fleiss intraclass correlation coefficients. Interobserver agreement was assessed by Kappa statistics and the Kruskal-Wallis test.
RESULTS. Fine anatomic detail was best presented in 0.5-mm or
1.00-mm reconstructions based on a 2.0 x 0.5 mm acquisition. A rotation
time of 
0.75 sec resulted in fewer artifacts; a significant dose
reduction was achieved with 80 kV and 100 mA at the expense of somewhat
increased noise, but without significant loss of anatomic detail in bone
presentation. Artifacts were tolerable with a pitch of 1.5 or less.
CONCLUSION. MDCT at the described optimal settings allows significant dosage reduction without sacrificing image quality. An acquisition and reconstruction thickness of 0.5 mm results in the best depiction of anatomic detail. A reconstruction thickness of 1.0 mm with a reconstruction interval of 0.5 mm represents a good trade-off between noise and resolution when using low-dose protocols.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
For musculoskeletal imaging, high speed and long scan length are of lesser importance [4, 5]. Small joints, such as the wrist, present a fine and sometimes challenging anatomy [6, 7]. For the clinician, the ability to obtain an excellent anatomic depiction in multiplanar reconstructions devoid of artifacts is of paramount importance. Although radiation exposure is of only minor importance in wrist imaging, the potential for dosage reduction in a high-contrast application should be exploited to obtain the best trade-off between diagnostic image quality and low radiation exposure [8, 9].
In the present study, we evaluated MDCT protocols on anatomic wrist specimens to determine their accuracy of anatomic presentation and to search for the balance of scan parameters that optimizes diagnostic image quality while keeping radiation exposure low.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
MDCT Protocol
All wrists were examined with a 4-MDCT scanner (Somatom Volume Zoom,
Siemens Medical Solutions).
In a first session, collimation and reconstruction slice thicknesses for presentation of wrist anatomy were evaluated. All wrists were scanned axially using collimations of 4 x 1 mm and 2 x 0.5 mm; a tube current of 130 mA; a voltage of 120 kV; a volume pitch of 1.5; a field of view of 12.8 cm, matching the minimal technical resolution of the scanner; and a scan length of 12 cm [10, 11]. Using the head of the capitate bone as a center of reference, we reconstructed coronal slices (60% overlap) [12-14] with slice thicknesses of 1.0 and 2.0 mm based on an acquisition slice thickness of 1.0 mm and of 0.5, 1.0, and 2.0 mm based on an acquisition slice thickness of 0.5 mm. All images were reconstructed by means of the high-resolution kernel (B70) and the soft-tissue kernel (B20) with a slice overlap of 60%. For coronal multiplanar reconstructions, both kernels were used. In addition, a high-field MRI examination was performed on all wrists using a Magnetom Vision 1.5-T scanner (Siemens Medical Solutions) with a four-channel phased-array wrist coil (Siemens Medical Solutions) and the following sequences: 2D coronal turbo STIR (TR/TE, 4,152/60; inversion time, 170 msec; slice thickness, 3.0 mm), 2D coronal spin-echo (450/12; slice thickness, 3.0 mm), and 2D coronal gradient-recalled echo fast low-angle shot with frequency-selective fat saturation (17.2/6.6; flip angle, 25°; slice thickness, 3.0 mm), with an in-plane resolution of 0.4 x 0.4 mm. Completing the protocol were two additional sequences: 3D axial dual-echo in the steady state with water excitation (43.7/9; flip angle, 35°; isotropic voxel size, 1 mm) and 3D coronal constructive interference in the steady state (12.3/5.9; flip angle, 40°; isotropic voxel size, 1 mm). Overall, the MRI protocol lasted about 20 min [15].
On the basis of the results of the first session, the pitch, rotation time, tube current, and voltage were adjusted to a collimation of 2.0 x 0.5 mm and reconstruction slice thicknesses of 0.5 mm (axial) and 1.0 mm (coronal). Different parameter settings for rotation time (0.5, 0.75, and 1.0 sec [pitch, 1.5; 120 kV; 130 mA]), volume pitch (1.0, 1.3, 1.5, and 2.0 [120 kV; 130 mAs; rotation time, 1.0 sec]), voltage (80, 120, and 140 kV [pitch, 1.5; 130 mA; rotation time, 1.0 sec]), and current (80, 100, 130, 160, and 200 mA [pitch, 1.5; 120 kV; rotation time, 1.0 sec]) were tested.
All images were reconstructed in axial and coronal planes using a high-resolution kernel (B70). Radiation dosage was estimated by application of the weighted CT dosage index as given by the manufacturer of the CT scanner [16, 17].
Compared with a collimation of 4.0 x 1.0 mm, the duration of a 2.0 x 0.5 mm scan doubles from 32 to 62 sec for a typical scan length of 12 cm using a pitch of 1.5 and a rotation time of one per second. Variation of pitch prolongs the scanning time linearly from 95 sec for a pitch of 1.0 to 62 sec for a pitch of 1.5 or 47 sec for a pitch of 2.0 (collimation for this example, 0.5 mm; rotation time, one per second). A reduction of rotation time results only theoretically in a linear shortening of the scanning time: The preparation time for the scan almost doubles for the preparation of two per second compared with one per second, and timesavings are only about one third.
Reading and Evaluation Protocol
All wrists were evaluated by three experienced observers in independent
sessions. A subset of 10 wrists was used to train the observers to use the
rating scales with focus on MR images and wrist anatomy. The independent
interpretations were performed 2 months later.
For all evaluations, soft copies were interpreted with bright monitors (260 candela, Siemens Medical Solutions) on a standard PACS reporting station (MagicView 1000, Siemens Medical Solutions). Fifteen criteria were used to evaluate gross anatomy and fine anatomic detail, and a general score was given for overall anatomic impression and for technical artifacts [14, 18-20]. For better evaluation of coronal reconstructions, spongy and cortical bone was examined in-plane along the x- and y-axes and along the table feed (z-axis).
For gross carpal anatomy, we evaluated trabecular structure in-plane (x- and y-axes, axial images), trabecular structure along table feed (z-axis, coronal images), trabecular outline in-plane (x- and y-axes, axial images), trabecular outline along table feed (z-axis, coronal images), cortical thickness in-plane (x- and y-axes, axial images), cortical thickness along table feed (z-axis, coronal images), smoothness of cortical outline in-plane (x- and y-axes, axial images), and smoothness of cortical outline along table feed (z-axis, coronal images). For fine carpal anatomy, we evaluated cartilage, junctional zone of cartilage and bone, insertions of radioscapholunate ligament, intrinsic capitate-to-hamate ligament, nutritive canals of lunate bone, nutritive canals of capitate bone, and conspicuity of ganglia and bone cysts. Other general evaluation criteria included technical artifacts not related to parameter settings and overall anatomic impression.
Direct comparison of different reconstruction modes and correlation with MRI within the same session was encouraged. A user-defined 5-point ordinal rating scale was used, in which a score of 1 (insufficient identification of anatomic structure) or 2 (probable identification of anatomic structure) indicated an image quality unsuitable for reporting and a score of 3 (sufficient quality of anatomic presentation), 4 (very good anatomic presentation), or 5 (optimal anatomic presentation), an image quality within diagnostic requirements.
The evaluation focused on both the gross anatomic delineation of bony structures and the fine anatomic detail of the carpal bones [21-23]. Soft tissue was not evaluated, because small air inclusions in the soft tissues of a subset of the anatomic specimens imparted a higher conspicuity to the triangular disk, for example, than normally occurs in vivo and because of the clear superiority of MRI in revealing these tissues.
Standard bone windows with a center of 200 H and a width of 2,000 H, or a center of 400 H and a width of 4,000 H, were recommended to the observers, but the window and zoom functions could be adjusted at any time during the interpretation and free use of these functions was encouraged. For easier identification of wrist anatomy and as a multimodal anatomic reference, the high-field MR images of the same 30 wrists were also available on the same PACS viewing workstations. The wrists were viewed by three observers in random order during separate sessions. The observers were un-aware of the acquisition parameters.
Biostatistics
All data were subjected to nonparametric testing. Shrout-Fleiss intraclass
correlation coefficients [24]
were used to assess intraobserver reliability. Kappa statistics and the
Kruskal-Wallis test were used to calculate and test interobserver agreement
(Table 1).
|
Kappa coefficients were calculated and tested for each variable of interest for each technique for all reviewers, because all three reviewers will occasionally agree by chance alone, even if they all are assigning ratings randomly.
Similarly, intraclass correlation was calculated and tested using the Shrout-Fleiss method [24] to assess interobserver reliability for each variable for each technique for all reviewers. Reliability—that is, the intraclass correlation—is essentially the extent of agreement between repeated measurements. Reliability factor values assumably always lie between 0 and 1, where 0 means not reliable at all and 1 means highly reliable. In other words, intraclass correlation is a measure to determine the similarity of the observations of one reviewer relative to the observations of a different reviewer.
Criteria of gross and fine anatomy were summarized and evaluated separately, and the results are presented as the medians and ranges for all evaluation series (Table 1).
Box-and-whisker plots illustrate the confidence intervals of the medians, quartiles, and 1.5 inter-quartile ranges of average scores (Fig. 1). All statistical analyses were performed using SAS statistical software, version 8.2 (SAS Institute), and all results were tested at a 5% level of significance.
|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Radiation Dosage
Doubling the volume pitch from 1.0 to 2.0 decreases dosage only to two
thirds of the original dosage (Table
2); half the dosage is reached using a pitch of 3. Reduction of
rotation time is, by comparison, effective: If the rotation time is increased
from 0.5 to 1.0 sec, weighted CT dosage index more than doubles. Tube voltage
is by far the most influential parameter on dosage: Although a similar
reduction of current is accompanied only by an almost linear saving of applied
dosage, lowering the voltage by approximately 40%, from 140 to 80 kV,
decreases weighted CT dosage index by more than 75%
(Table 3).
|
|
Acquisition and Reconstruction Slice Thicknesses
Kappa correlation coefficients for agreement between observers were high
for the first evaluation sessions focusing on acquisition and reconstruction
slice thicknesses (Table 1) and
varied little between the five parameter settings. No significant
interobserver differences were found using the Kruskal-Wallis test. The
Shrout-Fleiss test revealed high intraobserver confidence in the
identification of anatomic structures.
Average readings were higher for 0.5-mm acquisitions and for 0.5- or 1.0-mm reconstructions; all readings were within the diagnostic range (Fig. 1A). Most reconstructions based on 1.0-mm acquisitions were rated within the diagnostic range if reconstructed to 1.0-mm thickness and less than the diagnostic range when reconstructed to a 2.0-mm thickness. However, the differences between 0.5- and 1.0-mm-thick reconstructions based on 0.5-mm acquisitions were not significant.
Rotation Time and Volume Pitch
Interobserver agreement and confidence according to the Kruskal-Wallis test
were best for rotation times of 0.75 and 1.00 sec. Interobserver disagreement
was noted for a rotation time of 0.5 sec (p = 0.01). Average readings
were also lower for a rotation time of 0.5 sec, but all readings were in the
diagnostic range (Fig. 1B). The
reason for these lower ratings was seen in the fine anatomic detail and in
minor step artifacts in the coronal reconstructions. Rotation time had only a
minor influence on diagnostic image quality. In the evaluation of volume
pitch, interobserver agreement and confidence were also similar.
Average readings were lower for a pitch of 1.5 than for a pitch of 1.0 or 1.3 and remained within the diagnostic range, whereas a pitch of 2.0 showed a major difference from a pitch of 1.5 or smaller (Fig. 1C). No interreviewer difference was seen for volume pitches ranging from 1.0 to 1.5, whereas a significant difference was seen for a pitch of 2.0 (Kruskal-Wallis test, p = 0.01).
Tube Voltage and Current
For all voltages tested and a current of 100 mA or more, interobserver
agreement and confidence were good (Table
1). A voltage of 80 kV received higher average ratings than did
voltages of 120 or 140 kV, but ratings for all three sets were similar
(Fig. 1D). Because of the high
noise, the 80-mA current received the lowest ratings, although like the other
readings they remained within the diagnostic range
(Fig. 1E); the interobserver
difference was significant (Kruskal-Wallis test, p = 0.01).
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
|
|
|
|
The wrist represents a high-contrast structure with a great potential for dosage reduction.
Lowering tube voltage increases the attenuation effect of bone. In the present study, however, we observed only a minor improvement in bone contrast when voltage was reduced from 120 to 80 kV, with basically un-altered presentation of bone structure. Although the expectations for a higher bone contrast in the wrist may therefore be of minor importance, the potential of low-voltage techniques allowing dose reduction without impairing the signal-to-noise ratio should be stressed. In the present study, an increase of tube voltage from 120 to 140 kV did not notably improve image quality. A reduction of tube voltage to 80 kV, however, produced a slight increase in image contrast, which became conspicuous when a wider presentation window was used (Fig. 3): The contrast differentiation of spongy bone improved slightly, whereas the presentation of cortical bone remained unchanged. Nevertheless, the altered impression the image gives may be unfamiliar to some radiologists, but they can become accustomed to it quickly. A decrease in voltage allows a greater dosage reduction than does a proportional decrease in tube current [30, 31]. This potential to reduce dosage should be used even if the lower voltage may have to be compensated for partially in some patients by an increase in tube current [32].
When the tube current in the present study was reduced to 100 mA, image noise increased mainly in the periarticular soft tissues. Even fine anatomic detail was still presented with diagnostic quality. A tube current of 80 mA was rated much lower, indicating that the noise volume passed a threshold beyond which the image quality was unacceptable for reporting. Tube current can therefore be reduced to as low as 100 mA without a no-table loss of image quality. This is of course true only if the wrist is positioned cranial to the head and not at the side of the body.
A low voltage setting of 80 kV combined with a current of 100 mA allows for dosage reduction without loss of diagnostic image quality. We recommend using a low voltage in all wrist protocols. Any attenuation of radiation that is due to fiberglass or plaster casting should be compensated for by an increase in tube current but not in voltage.
A shortening of rotation time results in an almost linear decrease in dosage. Little difference is seen between rotation times of 0.75 and 1.00 sec, but minor step artifacts degrade the image quality when a rotation time of 0.5 sec is used, because the jumping focus is available only for a rotation time of 0.75 sec or more (Fig. 4). Thus, a rotation time of 0.75 sec is the optimal trade-off for quality and radiation dose. In imaging of the wrist, a volume pitch of 1.5 commonly is used for single-detector scanners. We tested a variety of pitches and found that the same 1.5 volume pitch is best for multidetectors as well. Figure 5A illustrates typical spiral artifacts using pitches of 1.0 and 3.0: Spiral artifacts project in different angular locations and are more pronounced using a larger pitch and on dense structures, such as the hydroxylapatite equivalent of the osteodensitometry phantom. If fine anatomic detail is to be examined and the reconstructions must be 0.5 mm, a smaller pitch of 1.3 reduces the step artifacts in some patients (Fig. 5B). These 0.5-mm thin reconstructions depict fine anatomic detail, such as the nutritive channels in the lunate and capitate bones, with a conspicuity much better than is shown by thicker reconstructions (Fig. 5B). A pitch smaller than 1.5 does not make much sense if the reconstruction thickness is 1.0 mm, because observers can perceive a difference only on reconstructions that are thinner than the volume pitch (Fig. 5C). The scanner software automatically adjusts the tube current to the pitch to keep both image noise and radiation dose constant when the pitch is changed. Although this automation simplifies use of the system, a rotation time of 0.75 sec or more and a pitch of 1.5 or less is still a reasonable parameter setting.
|
|
Although focusing on the anatomic presentation of the carpal bones, the present study could not assess the performance of its protocols on small abnormalities of the carpal bones. Instead, it focused on subtle anatomic details, such as nutritional channels and degenerative ganglia. In small and distinct carpal abnormalities such as nondisplaced fractures, distortion of the structure of the spongy bone and of minute anatomy (e.g., nutritional channels) is often an indirect sign of abnormality. We encourage future studies that focus on small and radiographically occult bone abnormalities.
In conclusion, wrist anatomy is best presented using an acquisition thickness of 0.5 mm on the 4-MDCT scanner evaluated. Fine anatomy is shown conspicuously with a reconstruction slice thickness of 0.5 mm, but 1.0-mm-thick reconstructions are acceptable in most cases. A pitch of 1.5 and a rotation time of at least 0.75 sec offer the optimal trade-off between radiation exposure and image quality.
A low voltage of 80 kV is recommended because it allows a large reduction in radiation dosage combined with a minor improvement in bone contrast. Image noise can be reduced by increasing tube current and leaving the low voltage settings unchanged. It is reasonable, in our opinion, to allow automatic adjustments by the scanner software. Even at low voltage, a tube current of as low as 100 mA is sufficient to obtain images of diagnostic quality if the wrist can be positioned cranial to the head.
Acknowledgments
Parts of this publication were taken from the doctoral thesis of Stefan
Galiano. We thank James Heywood for the review of the English language.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
This article has been cited by other articles:
|
|
 |
|
  T. Murase, K. Oka, H. Moritomo, A. Goto, H. Yoshikawa, and K. Sugamoto Three-Dimensional Corrective Osteotomy of Malunited Fractures of the Upper Extremity with Use of a Computer Simulation System J. Bone Joint Surg. Am., November 1, 2008; 90(11): 2375 - 2389. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 N. C. Dalrymple, S. R. Prasad, F. M. El-Merhi, and K. N. Chintapalli Price of Isotropy in Multidetector CT RadioGraphics, January 1, 2007; 27(1): 49 - 62. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
