| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Gastrointestinal Imaging |
1.5 cm) Liver Metastases: Is Thinner Collimation Better?1
1 From the Departments of Medical Imaging (M.A.H., M.M.A., D.C.R., M.E.O., A.E.H., G.A.L.) and Laboratory Medicine and Pathology (M.R.), and the University Avenue Hepatobiliary-Pancreatic Surgical Oncology Program (S.G.), Princess Margaret Hospital, University Health Network and Mount Sinai Hospital, University of Toronto, 610 University Ave, Toronto, Ontario, Canada M5G 2M9. From the 2000 RSNA scientific assembly. Received July 19, 2001; revision requested August 16; final revision received March 26, 2002; accepted April 18. Address correspondence to M.A.H. (e-mail: m.haider@utoronto.ca).
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
MATERIALS AND METHODS: Thirty-one patients underwent contrast material–enhanced multi–detector row helical CT before hepatic resection in this prospective study. Images were reconstructed at collimations of 5.00, 3.75, and 2.50 mm with 50% overlap and reviewed independently by three radiologists. Each lesion was characterized as metastatic, benign, or equivocal and graded for conspicuity. Criterion standards were pathologic assessment of the resected liver and follow-up of the nonresected liver. Only lesions 1.5 cm or smaller were analyzed.
RESULTS: There were a total of 88 liver lesions 1.5 cm or smaller, and 25 of these were metastases. Pooled sensitivity for all lesions improved with thinner collimation (66% [58 of 88 lesions], 69% [61 of 88], and 82% [72 of 88] at collimations of 5.00, 3.75, and 2.50 mm, respectively), and this was statistically significant (P = .01). However, no significant difference was noted between collimations in the pooled sensitivity for metastatic lesions (80% [20 of 25 lesions] at all collimations) (P > .99). No statistical difference was noted in the conspicuity of lesions at different collimations (P = .18).
CONCLUSION: Image reconstruction with multi–detector row helical CT at collimations less than 5 mm may not improve sensitivity in the detection of hepatic metastases 1.5 cm or smaller.
© RSNA, 2002
Index terms: Computed tomography (CT), helical, 761.12115 • Liver, CT, 761.12115 • Liver, surgery, 761.1267 • Liver neoplasms, metastases, 761.3317
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
The development of helical CT has allowed for faster scanning of the liver at thin collimations (5–7 mm) in a single breath hold during the portal venous phase of contrast material enhancement. This has resulted in improved sensitivity in the detection of metastases compared with that of conventional CT, with sensitivities reported between 74% and 85% (8,9). In these series, almost all the false-negative results involved lesions smaller than 1.5 cm in diameter. It would seem logical that sensitivity could be further improved if one were able to obtain thinner collimation scans through the liver.
The development of multi–detector row helical CT allows multiple images to be obtained simultaneously during spiral acquisition (10). Combined with improved x-ray tube performance, this technologic advance has allowed effective scanning of the liver at collimations less than 5 mm in a single breath hold.
The purpose of this prospective study was to determine if multi–detector row helical CT performed at collimations less than 5 mm would improve the detection of hepatic metastases 1.5 cm or smaller in diameter.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Any patient being considered for hepatic resection at our institution was considered eligible for the study. Only patients who actually underwent hepatic resection were included in the final analysis. Patients who at the time of surgery had extrahepatic disease that was not detected preoperatively were excluded.
Patient Population
A total of 43 patients who were being considered for hepatic resection were enrolled from March 11, 1999, to July 31, 2000. Eight patients underwent exploratory surgery and had disease that was too extensive for resection owing to peritoneal metastases, metastatic lymph nodes, or unexpected findings in size or number of hepatic lesions. Four patients were lost to follow-up. Our final group consisted of 31 patients, 30 with metastatic colorectal carcinoma and one with metastases from a gastrointestinal stromal tumor. The patient ages ranged from 42 to 82 years (mean, 66 years; median, 69 years). Twenty-two patients were male, and nine were female. The mean time between multi–detector row helical CT and surgery was 38.6 days (range, 7–90 days; median, 32 days). The mean time to follow-up CT was 216.9 days (range, 122–424 days; median, 189 days).
CT Technique
Scans were obtained by using a multi–detector row helical CT system (QX/i Lightspeed; GE Medical Systems, Milwaukee, Wis) with the following parameters: 120 kV, 230–330 mA, table speed of 7.5 mm per rotation, high-quality mode, and pitch of 3:1. Nonionic intravenous contrast material (iohexol [Omnipaque 300], 30 mg of iodine per milliliter; Amersham Health, Buckinghamshire, United Kingdom) was administered with the use of a power injector (Medrad, Indianola, Pa) at a dose of 2 mL/kg up to a maximum of 200 mL, at a rate of 3 mL/sec with a 60-second delay. The images were retrospectively reconstructed at collimations of 5.00, 3.75, and 2.50 mm with 50% overlap. Before injection of contrast material, a single 5-mm-collimation image was obtained through the liver with the same kilovoltage and milliampere settings.
Image Interpretation and Analysis
Images obtained at each collimation for each patient were presented in a randomized fashion to three experienced abdominal radiologists (D.C.R., M.E.O., A.E.H.) and reviewed in stack mode on a picture archiving and communication system (PACS) workstation (eFilm Medical, Toronto, Ontario, Canada). All studies were interpreted by each reader independently. The readers were blinded to the surgical or pathologic results. Technical parameters and patient identifiers were hidden from the reviewers at the time of interpretation. The radiologist was allowed to choose the window width and level for each series as he saw fit. For each finding, the radiologist was asked to note the largest dimension and characterize the finding as a metastasis, equivocal, or benign. The radiologists were given the following guidelines for classifying findings as benign: hypoattenuating, homogeneous, water attenuation, well-defined margins (ie, cyst); hypoattenuating, angular margins, typical location (ie, fatty infiltration); mixed hypo- and hyperattenuating with peripheral nodular enhancement (ie, hemangioma); and hyperattenuating, angular margins, typical location (ie, fatty sparing, flow phenomenon). All other findings were considered equivocal or malignant on the basis of the radiologist’s subjective interpretation. A conspicuity value of 1, barely visible; 2, adequate visibility; 3, good visibility; or 4, excellent visibility was assigned to each finding. If the total number of findings exceeded 10, then no further lesions were counted. Because the readers reviewed independently, differences in the segment localization and numbering of findings had to be resolved between the readers. Two radiologists (M.A.H., M.M.A.) who were not involved in image interpretation together reviewed all CT images and assigned each finding a specific hepatic segment and matched the lesion numbers between readers. If a finding bridged more than one segment, the finding was assigned the segment that it predominantly occupied. Thus, each finding was localized to a single segment. Lesion size was calculated by averaging all CT measurements for a given lesion in a given patient over all collimations with which the lesion was seen by a radiologist. Only lesions 1.5 cm or smaller were included in the final analysis.
Data analysis was performed to assess lesion detection, characterization, and conspicuity. Sensitivity was used as a measure of lesion detection. The radiologist’s characterization of a finding as benign, malignant, or equivocal was not taken into account for the calculation of sensitivity. Sensitivities were calculated on a lesion-by-lesion basis for each reader at each collimation for all lesions and for malignant lesions alone. When a reader made a finding (whether characterized as benign, equivocal, or malignant) and a corresponding lesion was present, this was a true-positive finding for any lesion. When a reader made a finding (whether characterized as benign, equivocal, or malignant) and a metastatic lesion was present, this was a true-positive finding for metastases. To eliminate the issue of interobserver variability and give the best possible scenario for sensitivity, an additional analysis of sensitivity was performed by pooling the results of all readers. In this case, if a true-positive interpretation was made by at least one reader it was considered a true-positive finding.
To give an idea of the radiologists’ ability to characterize lesions, counts were tabulated for the number of equivocal findings, the number of findings made when there was only normal liver tissue present (overcalls), and the number of misclassified findings. A finding misclassified as malignant was defined as one classified as malignant when in fact either the lesion was benign or normal liver was present. A finding misclassified as benign was defined as one classified as benign when a malignant lesion was present. Equivocal findings were not considered to be misclassified. Because of the small number of lesions, statistical analysis was not performed on these counts.
To assess conspicuity of radiologic findings, the mean conspicuity of lesions was calculated and compared for each reader at each collimation.
To assess noise and enhancement, two region-of-interest measurements were performed on a normal portion of the liver at each collimation for each patient and on the nonenhanced scan. All measurements were performed by one of the authors (M.M.A.). Care was taken to ensure that the diameter of the region of interest was greater than 1 cm, the area of the liver being interrogated was homogeneous, vessels were avoided, and the same region of the liver was measured across studies in the same patient. The mean Hounsfield unit measurement and SD were recorded. The SDs on the contrast material–enhanced studies were used as a measure of noise for each collimation (11). Liver enhancement was calculated by subtracting the mean contrast-enhanced region-of-interest measurements from the mean nonenhanced region-of-interest measurements.
Criterion Standards
The criterion standard was histopathologic assessment of the resected liver. The liver was sectioned in 8–10-mm slices and reviewed by means of direct visual inspection, with a radiologist (M.M.A., M.A.H.) and pathology technician present, and the results were compared with the imaging findings. The radiologist present at pathologic sectioning was one of the radiologists who also performed matching of lesions between readers. For cases in which a lesion was seen at CT and not at visualization of the gross pathologic specimen, thinner cuts were performed at 5 mm in an attempt to find the lesion. All suspicious areas were identified and sent for histologic assessment by the pathologist (M.R.). Intraoperative biopsy of any hepatic lesion was also considered a criterion standard. Follow-up CT more than 4 months after surgery was used as the criterion standard in the nonresected liver. An increase in size of a lesion was considered proof of metastases, and disappearance or stability was proof of benignity. When a hepatic lesion was present on the follow-up CT scan but was not depicted on the preoperative CT scan by any of the readers, the preoperative scans were reviewed again. If the new lesion was visible in retrospect on scans of any collimation, it was counted as a missed metastasis. Thus, new lesions were not counted unless retrospective review of the preoperative CT scans showed that the lesions were actually present but had not been identified prospectively. Intraoperative US and magnetic resonance (MR) imaging were considered the criterion standards for cysts in both resected and nonresected livers.
Statistical Methods
Sensitivity of lesion detection was calculated for each reader at each collimation as the proportion of true-positive findings to the total number of lesions. Similarly, the sensitivity after pooling the readers’ results was the proportion of interpretations at which at least one reader had identified a true-positive finding to the total number of lesions. Ninety-five percent CIs were calculated by using the Rao and Scott method for clustered data because there were multiple lesions in some subjects (12).
Sensitivity was tested to see if it differed between readers or collimations, after adjustments for the variation due to the other variable, by using conditional logistic regression (13) with the lesion number as the stratum variable. A binary variable indicating whether the lesion was detected was the outcome variable, and reader and collimation were the independent variables. Any potential dependence due to multiple lesions in a subject was removed by the conditioning.
Sensitivity of metastases detection, defined as the proportion of true-positive findings to the total number of metastases, was examined in the same manner.
Means and SDs of the conspicuity ratings were calculated for each reader at each collimation. We tested conspicuity to see if it differed across collimations after adjusting for reader by using a linear mixed model (14). The conspicuity rating was the outcome variable, and reader and collimation were the independent variables. Lesion number nested within subject identifier was included as a random effect to adjust for any correlation between the ratings. Least square means were calculated to estimate the mean conspicuity at each collimation after adjustments for the effect of the reader.
Interreader agreement was assessed for each pair of readers at the three different collimations by using 
statistics. The values between 0.40 and 0.75 are considered to represent fair to good agreement beyond chance (15).
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
There were a total of 88 liver lesions 1.5 cm or smaller in 31 patients. Of the 88 small lesions, 25 were metastases seen in 12 patients. The criterion standard was pathologic examination for 38 lesions and imaging for 50. Of the 50 lesions assessed with imaging, the criterion standard was follow-up in 41, MR imaging in two, and intraoperative US in seven. For the metastases, the criterion standard was pathologic assessment in all 25. Mean lesion size was 0.7 cm, with a range of 0.1–1.5 cm.
When we evaluated the results for each reader separately, sensitivity for detection of all small lesions at the three collimations was 48%–65% (42–57 of 88 lesions), and with pooled results from all readers, sensitivity was 66%–82% (58–72 of 88 lesions) (Table 1). Overall, for the three readers, a significant difference in sensitivity at different collimations was not detected (P = .06). There was a significant difference in sensitivity between readers (P = .03). Readers B and C both showed a trend toward increased sensitivity at thinner collimations. When looking at the pooled results, there was a significant improvement in sensitivity at thinner collimation (P = .01).
|
|
|
|
|
values from 0.31 to 0.57.
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
The advent of helical CT has resulted in sensitivities that are significantly higher than that of conventional CT and in false-positive rates that are lower than that of CT during arterial portography; however, sensitivities are still below 90%, especially for small lesions (9,18). Multi–detector row helical CT represents an important technologic advance over single–detector row helical CT. With the development of quadsection technology and an increase in gantry speed to two revolutions per second, multi–detector row helical CT is up to eight times faster than conventional single–detector row helical CT (19). The decrease in acquisition time allows for thinner collimation during the portal venous phase of hepatic enhancement (10). In addition, at the expense of increased noise, images can be retrospectively reconstructed at progressively thinner collimations depending on the table speed and pitch used. It would seem logical that higher spatial resolution would lead to improved sensitivity for the detection of liver metastases. We are unaware of any other published study with pathologic assessment as a criterion standard to evaluate multi–detector row helical CT in this setting.
Our results showed significantly increased pooled sensitivity for small-lesion detection at thinner collimation; however, when analysis of metastases alone was performed there was no demonstrable improvement in sensitivity with thinner collimation. It is possible that contrast material delivery mechanisms may play a larger role than spatial resolution in the detection of small metastases. Use of contrast material doses based on patient body weight and imaging timed to peak hepatic enhancement are advocated to achieve improved lesion detection by maximizing the contrast between the enhanced liver and the relatively nonenhanced metastases (20). Peak enhancement of 50 HU has been suggested as necessary for tumor detection; however, this number may be somewhat arbitrary (21). CT during arterial portography with peak hepatic enhancement on the order of 100 HU (22) has higher sensitivity than that of helical CT for detection of metastases and in particular for small metastases. Overall sensitivities for detection of hepatic metastases of 93%–100% have been reported with CT during arterial portography (5–7,17,23,24) compared with no more than 85% for helical CT (9,25). Sensitivities for detection of small lesions are much lower. Schmidt et al (26) published a series comparing helical CT at 5-mm collimation with helical CT during arterial portography and showed that the sensitivity for the latter was 100% compared with 17% for helical CT for detection of metastases smaller than 1 cm in diameter. Others have reported the sensitivity of helical CT for detection of hepatic metastases smaller than 1 cm to be 56% (25). In a study by Valls et al (9), all false-negative findings were for metastases smaller than 1.5 cm. Perhaps peak liver enhancement that is higher than that achieved with currently accepted CT protocols may be required for the detection of small metastases, and this factor may be more important than section thickness.
Study Limitations
Noise has an important role in lesion conspicuity and detection. It is possible that decreasing noise at thinner collimation might improve detection of metastases. Weg et al (27), by using different CT technology (dual-section helical CT Elscint Twin, Hackensack, NJ), were able to achieve lower image noise at comparable collimations and reported noise values of 8.1 at 5 mm and 9.7 at 2.5 mm compared with 14.4 and 21.3, respectively, in our study. This may have been a factor in minimizing the benefit of thinner collimation for metastatic disease.
Interobserver agreement was suboptimal for the detection of small hepatic lesions. This may have been related to a number of factors, such as varied window widths or levels and scroll speeds on the workstations. We dealt with this limitation in two ways: first, by adjusting for interobserver variability in the modeling analysis and second, by using a pooled reader sensitivity to eliminate interobserver variability as an issue in sensitivity analysis.
Admittedly, longer patient follow-up may have revealed that there were more metastases in the nonresected liver than was shown in our results. Given that new liver metastases were not counted in this study, we suspect follow-up was not a major confounding factor, as the nonresected liver is carefully screened with intraoperative US in this patient population and biopsy is performed intraoperatively on all suspicious lesions. Thus, the likelihood of a preoperatively detected lesion not undergoing biopsy or not being definitively characterized before resection would be small.
In the pathologic assessment of the resected liver, a potential detection bias existed because we obtained thinner sections in areas where CT showed an abnormality but no lesion could be seen in the 8–10-mm pathologic section. This may have raised sensitivity for metastases and lowered rates for overcalls, but we expect this effect would be similar across all collimations.
A 50% overlap was used in image reconstruction. Although we recognize that this is not performed routinely at many institutions because of the advent of the filmless environment (i.e., PACS) and the known improvement in lesion detection provided by 50% overlap, we suggest that a 50% overlap should be a routine part of the assessment of patients in the preoperative setting (28). The overlapping reconstruction combined with the relatively lower noise level at 5-mm collimation may have improved the sensitivity and conspicuity of the thicker reconstructions; however, because we performed 50% reconstruction at all collimations it would not have affected the comparative analysis of sensitivity.
Effect on Clinical Practice
Eighty percent of liver lesions smaller than 1 cm in diameter in patients with a known primary malignancy are benign (29). Detecting more benign lesions while not detecting more malignant lesions at thinner collimation would not be of great clinical value. If one were to see more lesions, then being able to characterize them as benign or malignant would be important. In this study, we saw no trend toward increased characterization of lesions as benign or malignant at thinner collimation. Although we have insufficient numbers of patients to reach a definitive conclusion, this brings into question the benefit of collimation less than 5 mm for characterization of small liver metastases. The benefit of thinner collimations may not be realized unless image noise is minimized and contrast material delivery mechanisms are further improved. Consensus reading of studies may also be necessary since interobserver variability in small-lesion detection may be high. In addition, because of the low prevalence of metastasis smaller than 1.5 cm in diameter, further studies with larger populations may be necessary to show whether there is a benefit to using thinner collimation in this clinical setting.
The current practice of performing helical CT at 5-mm collimation with 50% overlap in the portal venous phase continues at our institution for all patients with metastatic colorectal carcinoma who are being assessed for consideration of potential liver resection. We still rely on US, MR imaging, and intraoperative US for addressing equivocal CT findings in candidates for hepatic resection. Multi–detector row helical CT has not fundamentally altered our imaging algorithm in this patient population.
In conclusion, routine reconstruction of images at collimations less than 5 mm with use of multi–detector row helical CT may not increase detection of small (
1.5 cm) liver metastases.
|
ACKNOWLEDGMENTS |
|---|
|
FOOTNOTES |
|---|
Author contributions: Guarantor of integrity of entire study, M.A.H.; study concepts, all authors; study design, M.A.H.; literature research, M.M.A., M.A.H.; clinical studies, M.A.H., M.M.A.; data acquisition, M.A.H., M.M.A.; data analysis/interpretation, M.A.H.; statistical analysis, G.A.L.; manuscript preparation, M.A.H.; manuscript definition of intellectual content, all authors; manuscript editing, M.A.H.; manuscript revision/review, M.A.H., M.E.O., G.A.L.; manuscript final version approval, all authors.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  R. Iezzi, A. R. Cotroneo, A. Filippone, F. Di Fabio, M. Santoro, and M. L. Storto MDCT Angiography in Abdominal Aortic Aneurysm Treated with Endovascular Repair: Diagnostic Impact of Slice Thickness on Detection of Endoleaks Am. J. Roentgenol., December 1, 2007; 189(6): 1414 - 1420. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
D.-M. Koh, G. Brown, Z. Meer, A. R. Norman, and J. E. Husband Diagnostic Accuracy of Rim and Segmental MRI Enhancement of Colorectal Hepatic Metastasis After Administration of Mangafodipir Trisodium Am. J. Roentgenol., February 1, 2007; 188(2): W154 - W161. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
P. T. Johnson and E. K. Fishman IV Contrast Selection for MDCT: Current Thoughts and Practice Am. J. Roentgenol., February 1, 2006; 186(2): 406 - 415. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
G. S. Sudakoff, D. P. Dunn, R. S. Hellman, M. A. Laguna, C. R. Wilson, R. W. Prost, D. C. Eastwood, and H. J. Lim Opacification of the Genitourinary Collecting System During MDCT Urography with Enhanced CT Digital Radiography: Nonsaline Versus Saline Bolus Am. J. Roentgenol., January 1, 2006; 186(1): 122 - 129. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 H. I. Khalil, S. A. Patterson, and D. M. Panicek Hepatic Lesions Deemed Too Small to Characterize at CT: Prevalence and Importance in Women with Breast Cancer Radiology, June 1, 2005; 235(3): 872 - 878. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
S. Saini Multi-Detector Row CT: Principles and Practice for Abdominal Applications Radiology, November 1, 2004; 233(2): 323 - 327. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 A. Kawashima, T. J. Vrtiska, A. J. LeRoy, R. P. Hartman, C. H. McCollough, and B. F. King Jr CT Urography RadioGraphics, October 1, 2004; 24(suppl_1): S35 - S54. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
E. M. Caoili, R. H. Cohan, A. Kawashima, and A. J. LeRoy Invited Commentary * Authors' Response RadioGraphics, October 1, 2004; 24(suppl_1): S55 - S58. [Full Text] [PDF]
|
|
|
|
|
|
A. Napoli, C. Catalano, G. Silecchia, P. Fabiano, F. Fraioli, F. Pediconi, F. Venditti, N. Basso, and R. Passariello Laparoscopic Splenectomy: Multi-Detector Row CT for Preoperative Evaluation Radiology, August 1, 2004; 232(2): 361 - 367. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
K. W. Kim, A. Y. Kim, T. K. Kim, S. H. Park, H. J. Kim, Y. K. Lee, M.-S. Park, H. K. Ha, P. N. Kim, J. C. Kim, et al. Small (l.e. 2cm) Hepatic Lesions in Colorectal Cancer Patients: Detection and Characterization on Mangafodipir Trisodium-Enhanced MRI Am. J. Roentgenol., May 1, 2004; 182(5): 1233 - 1240. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
