| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
1
Department of Radiology, Neuroradiology/Head and Neck Imaging, Roswell Park
Cancer Institute, Elm St. and Carlton St., Buffalo, NY 14263.
2
Department of Ultrasound, Roswell Park Cancer Institute, Buffalo, NY
14263.
3
Department of Neurosurgery, State University of New York at Buffalo, Millard
Fillmore Hospital, Gates Cir., Buffalo, NY 14263.
4
Department of Cancer Prevention, Epidemiology and Biostatistics, Roswell Park
Cancer Institute, Buffalo, NY 14263.
5
Department of Diagnostic Radiology, Boston Deaconess-Beth Israel Medical
Center, 330 Brookline Ave., Boston, MA 02115.
6
Department of Diagnostic Radiology, Brigham and Women's Hospital, 75 Francis
St., Boston, MA 02115.
Received January 7, 2000;
accepted after revision March 30, 2000.
Preliminary work presented at the annual meeting of the American Roentgen
Ray Society, Boston, May 1997.
Abstract
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
MATERIALS AND METHODS. Twenty-three patients with suspected intracranial abnormalities related to the brain base, as determined either by clinical examination or at the time of imaging, were evaluated with contrast-enhanced conventional CT of the brain (5-mm collimation, 140 kVp, 170 mA, 2-sec rotation time) and reformatted helical CT (1-mm collimation, 1.5 pitch, 120 kVp, 220 mA). Helical sections were reformatted to a thickness of 5 mm by a volume-averaging algorithm using a computer workstation. Three observers retrospectively and blindly reviewed the images and qualitatively scored artifacts at the foramen magnum, middle cranial fossa, anterior cranial fossa, interpetrous region, and internal occipital protuberance. Image graininess and observer confidence were also scored. Paired statistical analyses using score differences in each patient were possible.
RESULTS. Reformatted helical CT reduced skull base—related artifacts across all five anatomic regions (p < 0.05). The foramen magnum showed the greatest reduction in artifacts and the anterior cranial fossa the least. Image graininess was increased on reformatted CT compared with conventional CT (p < 0.05), but observer confidence remained higher for reformatted CT (p < 0.05). Total additional scan time was 3.15 ± 0.38 min with 5.3 ± 1.2 min required for reformatting.
CONCLUSION. Reformatted CT significantly decreases skull base—related artifacts in the brain, improving confidence in evaluation of the brain base and adding an average of only 8.45 ± 1.6 min of scanning and processing time to each examination.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Previous studies have used computer postprocessing for reduction of skull base—related artifacts on CT [1,2,3,4]. Other studies have used a prospective scanning algorithm and filter selection to improve images of the posterior fossa. These techniques require additional hardware or software for routine implementation that may not be readily available commercially. Selection of an optimal scan angle (5° below Reid's baseline) can also improve CT scans of the brain base without the need for new software or hardware, but significant artifacts remain a problem [5, 6].
There is evidence that nonlinear volume averaging and beam-hardening artifacts are major contributors to skull base—related artifacts on CT. Thin-beam collimation in combination with summation of the thin sections by retrospective reformation has effectively decreased these artifacts in phantom studies and volunteers [7]. Although this technique works with conventional axial CT, increased scan time and radiation dose are limiting factors that make it less practical.
Helical CT improves the quality of multiplanar reformatting and three-dimensional reconstructions by decreasing motion and misregistration artifacts. The speed of helical CT, compared with that of conventional CT, has made scanning with 1- or 2-mm collimation practical. Increased helical pitch allows an extended scanning range with thinner beam collimation without increasing radiation dose [8]. We have used this technique since July 1998 for evaluation of patients with suspected posterior fossa abnormalities. This is a retrospective analysis of all CT scans obtained with this technique at our institution from July 1998 through November 1999. We hypothesize that this technique decreases nonlinear volume-averaging artifacts by virtue of the thin collimation and decreases beam-hardening artifacts by summation of variable signal loss on the thick-section reformatted images.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Scanning
All patients were imaged on a helical scanner (HiSpeed Advantage; General
Electric Medical Systems, Milwaukee, WI) with the scan plane parallel to the
orbital roof. Conventional CT scans (140 kVp, 170 mA, 2-sec rotation time)
were generated with 5-mm collimation continuously from the skull base to the
vertex 75 sec after initiating a 95-mL IV bolus of a 300 mg/mL iopromide
solution (Ultravist; Berlex Laboratories, Wayne, NJ) at a rate of 1 mL/sec.
Helical scans of the brain base were generated with a 1.5 pitch, 1-mm
reconstruction interval, and 1-mm collimation (120 kVp, 220 mA) from the
foramen magnum to the top of the anterior clinoid process. All helical scans
were obtained after contrast material administration. When helical scans were
obtained first, contrast dose and delay were identical to that for the
conventional contrast-enhanced study. Helical scans were obtained immediately
before conventional contrast-enhanced CT in 11 patients and immediately after
conventional contrast-enhanced CT in 12 patients. Scan angle and field of view
were constant for conventional and helical studies. Image reconstruction was
determined by the scanner manufacturer and included filtered back projection
with an iterative bone processing filter for the axial image set (IBO filter;
General Electric Medical Systems) and an 180° linear interpolation for
helical section reconstruction. Each image reconstruction required 4 sec.
Printing and Reformatting
For the purpose of the study, all image data including conventional and
helical scans were restored to a workstation and reprinted using the same
processor, magnification, film format, and window and level settings (window =
100 H, level = 35 H). Patient data and scan annotation were also removed on
the reprinted images. Helical scans were reformatted to a 5-mm thickness with
5-mm intervals using a volume-averaging algorithm and a computer work-station
(Advantage Windows 1.2; General Electric Medical Systems). A nine-on-one
filming format was used. Average reformatting time was recorded for each
patient.
Observers
Three observers reviewed each of the 46 image sets. Observers consisted of
a neuroradiologist with a certificate of added qualification, a radiologist
with fellowship training in cross-sectional imaging, and a neurosurgeon.
Random numbers were assigned to each case, and patient demographic and
technical annotation were blocked from the films to keep the observers unaware
of patient name, scan type, and prior interpretation. Each observer scored
artifacts on a 5-point scale, with scores of 1 representing barely perceptible
artifacts, 2 representing noticeable artifacts that do not affect
interpretation, 3 representing artifacts that slightly limit interpretation, 4
representing artifacts that limit interpretation and require follow-up with MR
imaging, and 5 representing severe artifacts that render interpretation
impossible. Four anatomic locations were scored including the foramen magnum,
interpetrous region, middle cranial fossa, and anterior cranial fossa. Streak
artifacts from the internal occipital protuberance were also scored. Observer
confidence was scored on a 5-point scale, with scores of 1 representing
complete lack of confidence (do not bill for this examination), 2 representing
very low confidence (follow-up with MR imaging is required), 3 representing
low confidence (consider MR imaging if clinical suspicion is high), 4
representing high confidence (suggest MR imaging only if clinical findings
strongly disagree with CT findings), and 5 representing maximal confidence (it
is unlikely that MR imaging will provide additional information). Overall
image graininess was scored on the same 5-point scale as the region-specific
artifacts. Each artifact and confidence level score was specifically defined
to minimize interobserver variability. After initial review, the number of
lesions identified on each of the image sets was determined by consensus. The
observers remained unaware of patient name, scan type, and prior
interpretation during this phase of the study.
Radiation Dose
A measure of relative dose levels for the two techniques was obtained using
an ionization chamber. Each scan technique was repeated three times over an
identical 10-mm surface through the ionization chamber. The ratio of charge
produced in the chamber during each technique was used to calculate the
percentage of difference in radiation dose between the techniques. Absolute
doses were calculated on the basis of data from a calibrated CT dose phantom
(ionization chamber model c1152, Capintec, Ramsey, NJ; or electrometer model
35617, Keithley, Cleveland, OH) and expressed in centigrays.
Statistics
Data were initially recorded on a computer spreadsheet (Microsoft Excel
'97; Microsoft, Redmond, WA). Data transformations and manipulations were made
with a commercially available statistical analysis package (SigmaPlot;
Statistical Package for the Social Sciences, Chicago, IL). Statistical
analyses were performed in a statistical analysis package available through
Pennsylvania State University (MINITAB; Minitab, State College, PA), with
microcomputer workstations. Because each observer assessed both images from
each patient, one for each technique, the difference (reformatted CT versus
conventional CT) was calculated for each of the seven scores, and these seven
score differences were the endpoints used for subsequent analyses. The
normality of the distributions of the seven score differences was examined
with the Ryan-Joiner test [9],
and, even though the score differences are integer rank differences (an
ordinal variable), an assumption of normality was found to be reasonable. The
normality assumption justified the subsequent use of parametric statistical
procedures.
The influence of observer and site on the artifact score difference and the influence of observer on the graininess difference and on the confidence difference were examined with a general linear model procedure (a type of analysis of variance). Comparisons of the influence of site on artifact score were also made using the Tukey-Kramer multiple comparison procedure of MINITAB [10].
The mean time and standard deviation for reformatting and printing the images were calculated.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
|
|
|
|
|
There was no significant difference in scoring among the observers for any of the difference scores (p > 0.05) as determined by the general linear model procedure in MINITAB (a type of analysis of variance). Anatomic site was found to be a significant factor for artifact difference (p = 0.003). The mean plus or minus standard error for the five sites was interpetrous, -1.79 ± 0.77; occipital protuberance, -1.76 ± 0.77; foramen magnum, -2.09 ± 0.12; middle cranial fossa, -1.79 ± 0.094; and anterior cranial fossa, -1.54 ± 0.089. When individual site differences were analyzed with the Tukey-Kramer multiple comparisons procedure [10], the foramen magnum site showed a significantly greater improvement in artifacts with reformatted CT than the anterior cranial fossa site (p = 0.001).
The mean time for reformatting and printing the images was 5.3 ± 1.2 min per patient. All images were obtained within an 11-min interval (mean, 6.9 ± 2.55 min). On average, 41 ± 5 sections were obtained in each helical series, with a total average scan time of 27.33 ± 3.33 sec. Image reconstruction time was 4 sec per additional image, averaging 2.7 ± 0.33 min.
A total of 34 lesions were detected with the reformatted helical CT and 24 lesions with the conventional technique (Figs. 4A,4B,5A,5B,6A,6B,6C,6D). One lesion detected on reformatted helical CT and not seen on conventional CT was confirmed to be present on MR imaging (Fig. 6A,6B,6C,6D), and one lesion suspected on conventional CT but not seen on reformatted helical CT was confirmed to be absent on MR imaging.
|
|
|
|
|
|
|
|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Our purpose was to compare a simple technique for reducing skull
base—related artifacts on CT using thinly collimated images reformatted
to 5-mm thickness with conventional 5-mm-thick axial images. The combination
of thin collimation and retrospective reformatting should decrease skull
base—related artifacts by a combined effect of decreased nonlinear
volume averaging and decreased beam hardening. First, nonlinear volume
averaging between objects of different densities has been shown to decrease
with thin collimation [7].
Unfortunately, thin collimation decreases photon flux because of constraints
related to tube cooling and small focal spot size. As a result, image
graininess increases. When the data from the thinly collimated scans are
averaged, the effective photon flux is increased proportional to the number of
sections averaged. Thus the lack of photon flux is partially compensated. The
second advantage of averaging the thin sections is a decrease in
beam-hardening artifacts, which is inversely proportional to the product of
the square root of the number of sections averaged and the milliamperes
(Pn
[A / {NxQ}]0.5, where Pn = pixel noise,
A = attenuation coefficient, N = number of sections averaged, and Q =
radiation dose per section). These principles apply to thin sections obtained
with either conventional axial imaging or helical imaging. Helical technique
was chosen because it is faster, delivers less radiation, and is free of
misregistration artifacts. Measurement of relative radiation dose revealed a
54% decrease in dose for the helical CT as compared with the conventional CT
used in this study. We routinely use this technique in addition to
conventional CT for patients with suspected posterior fossa abnormalities or
with excessive skull base—related artifacts on conventional CT. We also
routinely use this technique for patients with symptoms related to the
posterior fossa who have contraindications for MR imaging.
Our study group consisted of 23 patients scanned with conventional and thin-section helical techniques. This group represented all patients scanned with this technique between July 1998 and November 1999 as determined by a review of scanning record books from our helical scanner. The helical scanning technique was selected because it could almost always be performed immediately after the conventional CT without the need for prolonged tube-cooling. Tube-cooling did result in a delay between the conventional and helical techniques in six of the 23 cases reviewed, with an average delay of 2 ± 0.56 min. This delay contributed to the average delay between conventional and helical scans but only occurred when the conventional scans were obtained first.
Because the image sets were paired, a difference score was used as the basis of all statistical comparisons. This technique uses each patient as his or her own internal control, thereby reducing data variance and increasing the power of the statistical procedures.
Our results support the hypothesis that reformatted CT significantly reduces skull base—related artifacts, relative to conventional CT (p < 0.05). The foramen magnum site had the highest decrease in artifacts. Although reformatted CT scans were significantly more grainy than conventional CT scans, all three reviewers had higher confidence levels in their interpretation of the reformatted CT scans. Previous studies have used variations in Hounsfield units across a region of interest as an objective measure of artifacts at the brain base [2, 6]. Unfortunately, the reformatted images could not be quantitatively assessed for pixel intensity variation on the basis of Hounsfield units. Software constraints would not allow these measurements after the images were reformatted, thus making quantitative analysis of Hounsfield unit variability in a given voxel impossible; another method of assessing artifacts was required. Yeoman et al. [5] used a subjective evaluation by two radiologists as a qualitative assessment of artifacts between two scanning techniques. We chose a defined qualitative assessment of artifacts and confidence level as described in the "Materials and Methods." The lack of significant variability among the qualitative artifacts and confidence scores among the observers, as evaluated by analysis of variance, supports the contention that the qualitative scores represent a true evaluation of the difference between the imaging techniques.
Some imaging features that may be affected by the helical technique were not evaluated in this study. Direct review of the helical source images with bone windows is likely to provide more detailed information about the skull-base anatomy and temporal bone than is available on conventional head CT. Thin-section helical source images can also provide the data needed for high-quality three-dimensional and multiplanar reconstruction of the bone. A detailed evaluation of part of the intracranial vasculature is also likely to be possible with the helical technique. It is possible that patient motion during the helical scan would significantly degrade the reformatted images. Fortunately, the helical scan time is short, and we did not encounter any motion-degraded images in our patient population. If motion occurred, review of the source images from the helical scan or repeating the helical series may represent a reasonable solution.
Disadvantages of the helical technique include the inability to obtain density measurements on the reformatted images. When Hounsfield unit measurements are necessary (for evaluation of an intracranial lesion for fat or fluid content or for enhancement), the source images can be assessed directly. Additional scan time and reconstruction time are also limitations of this technique. The addition of the thin-section CT added 6.9 min to the scanning time including an additional 5.3 min for reformatting the images at the workstation. When image reconstruction time (time for the raw data to be converted to an image) is considered (4 sec per image on our scanner), approximately 15 min of total additional time is needed for the helical technique. Additional data storage space was also required for the 41 ± 5 sections obtained in each helical series. A small increase in film cost resulted from an additional sheet needed to print the reformatted images. Although reformatting the images at the operator console or independent console is possible, it is less intuitive than the process on the workstation. Depending on the workstation and console used, reformatting may be faster or slower than the results we obtained.
Because this was a retrospective comparative analysis of artifacts between two CT techniques, the conventional axial CT was considered the gold standard in terms of artifact assessment. Lesion detection was recorded by a consensus of the observers. MR imaging correlation was available in seven cases within 1 month of the CT in this patient group. Retrospective comparison revealed one lesion detected with thin-section reformatted CT that was not seen with conventional techniques. In this case, reformatted helical CT was performed after conventional CT with a delay of 8 min. It is possible that the delay resulted in increased lesion conspicuity in the helical CT due to delayed contrast enhancement. In a second case, a lesion suspected on conventional CT was not present on MR imaging or thin-section reformatted CT. This comparison is not statistically valid but does suggest that the reduction in artifacts with the helical reformatted technique may result in an improved sensitivity and specificity for lesion detection. A formal prospective comparison with MR imaging as the gold standard would be an interesting subject of future research.
In conclusion, reformatted helical CT of the brain base is a simple, readily available technique for decreasing artifacts related to the skull base. Although reformatted helical CT images appear grainier than those of conventional CT, observer confidence was significantly higher when interpreting the reformatted helical CT. Evaluation of the foramen magnum and cervical—medullary junction is particularly improved with reformatted CT. Drawbacks of the technique include an inability to perform direct Hounsfield unit measurements on the reformatted images, additional data storage and scanning time requirements, and the need for image postprocessing. We continue to use this technique routinely to evaluate patients with clinically suspected brain base abnormalities or with suspicious areas on conventional CT scans.
Acknowledgments
We thank Roswell Park Memorial Cancer Institute for help with statistical
analysis of the data.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
This article has been cited by other articles:
|
|
 |
|
  K Imai, M Ikeda, S Wada, Y Enchi, and T Niimi Analysis of streak artefacts on CT images using statistics of extremes Br. J. Radiol., November 1, 2007; 80(959): 911 - 918. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. van Straten, H.W. Venema, C.B.L.M. Majoie, N.J.M. Freling, C.A. Grimbergen, and G.J. den Heeten Image Quality of Multisection CT of the Brain: Thickly Collimated Sequential Scanning versus Thinly Collimated Spiral Scanning with Image Combining AJNR Am. J. Neuroradiol., March 1, 2007; 28(3): 421 - 427. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
