ORIGINAL ARTICLE

Evaluation of Computed Tomography for Determining the Diagnosis of Acetabular Fractures

Robert V. OToole, MD, Garrick Cox, MD, K. Shanmuganathan, MD, Renan C. Castillo, MS, Clifford H. Turen, MD, Marcus F. Sciadini, MD, and Jason W. Nascone, MD
Key Words: acetabular fractures, computed tomography, plain radiographs, diagnosis, three dimensional reconstructions, simulated Judet views (J Orthop Trauma 2010;24:284290)

Objective: We assessed whether, in contrast to reports in the

literature, computed tomographic (CT) scans improve the ability to classify acetabular fractures in comparison with plain radiographs.

Design: Prospective. Setting: Level I trauma center. Patients: Seventy-ve patients with 75 acetabular fractures treated
between June 2005 and May 2006.

INTRODUCTION
Adequate radiographic assessment is essential for the diagnosis and treatment of acetabular fractures. Three plain radiographic views traditionally have been used to dene the fracture pattern: anteroposterior view, obturatoroblique Judet view, and iliacoblique Judet view of the pelvis.13 Once computed tomography came into common use, axial view computed tomographic (CT) scans were added to the Judet views for preoperative evaluation.4 Preoperative radiographic evaluation with the Letournel and Judet classication system typically is used to classify acetabular fractures18 and to plan for operative approaches. The Letournel and Judet system includes 10 fracture types that are divided into ve elementary fracture patterns and ve associated fracture patterns.3 One previous study analyzed the interobserver reliability of the Letournel and Judet fracture classication system and found substantial interobserver and intraobserver agreement with kappa values on the order of 0.7.5 Recent advances in CT include the capability to produce three-dimensional CT reconstructions and simulated anteroposterior and Judet view radiographs derived from CT scans.912 Despite the emerging use of the new threedimensional CT imaging modalities to help classify and plan for treatment of acetabular fractures, the inuence of the imaging modalities on diagnosis has not yet been well characterized. The few studies examining the issue have questioned the usefulness of CT scans for evaluating acetabular fractures.5,13 Our hypothesis was that the use of CT scans improves accuracy in classifying acetabular fractures in comparison with plain radiographs alone.

Intervention: Four different image sets for each patient were evaluated: image set A, Judet view plain radiographs plus axial view CT scans; image set B, Judet view plain radiographs alone; image set C, three-dimensional CT reconstructions; and image set D, CTsimulated anteroposterior and Judet views of the pelvis. The 300 image sets were viewed in random order by four orthopaedic trauma fellowship-trained surgeons who independently recorded a diagnosis. A gold standard diagnosis was determined by group consensus. Main Outcome Measurements: Agreement among four imaging methods was evaluated by using kappa statistics for multiple raters and nominal data. Results: Comparing the gold standard diagnosis with the four image sets, Judet view plain radiographs had a worse kappa value than CT scans (P , 0.05). The adjusted kappa values for all three image sets that included CT scans averaged greater than 0.62, showing substantial agreement, whereas the image set with plain radiographs alone (image set B) had a lower kappa value of only 0.48 (P , 0.05). Conclusions: In contrast to previous reports in the literature, the accuracy of plain radiographs alone was less than the accuracy of CT scans in terms of diagnosis. The interobserver reliability was also worse for plain radiographs alone.

Accepted for publication October 28, 2009. From the R Adams Cowley Shock Trauma Center, Department of Orthopaedics, University of Maryland School of Medicine, Baltimore, MD. The authors report no nancial disclosures related to the content of this manuscript. Reprints: Robert V. OToole, MD, 22 S. Greene Street, T3R62, R Adams Cowley Shock Trauma Center, Department of Orthopaedics, University of Maryland School of Medicine, Baltimore, MD 21201 (e-mail: rvo3@ yahoo.com). Copyright 2010 by Lippincott Williams & Wilkins

PATIENTS AND METHODS Inclusion Criteria

After obtaining Institutional Review Board approval, we retrospectively reviewed a database that had been prospectively collected and designed for this study. Between June
J Orthop Trauma  Volume 24, Number 5, May 2010

284

| www.jorthotrauma.com

J Orthop Trauma  Volume 24, Number 5, May 2010

Determining the Diagnosis of Acetabular Fractures

2005 and May 2006, 178 consecutive patients presented at our Level I trauma center with acetabular fractures. Patients were excluded from the study for the following reasons: if any of the types of images we were studying were not available for viewing; if the imaging studies had not been completed; and if any identiable markers such as associated fractures or hardware from previous treatment were seen on the images. Most of the exclusions were because three-dimensional CT reconstructions were not available for analysis. However, no patient was excluded because of poor image quality because we wanted to study the imaging modalities as they were used in clinical practice as opposed to including only ideal images that might be very difcult to routinely obtain. After exclusions, the study group included 75 patients with 75 fractures.

Image Sets
Four different image sets were evaluated for each patient: image set A, Judet view plain radiographs plus axial view CT scans; image set B, Judet view plain radiographs alone; image set C, three-dimensional CT reconstructions (Fig. 1); and image set D, CT-simulated anteroposterior and Judet views of the pelvis (Fig. 2). This created 300 total image sets for viewing. The 300 individual image sets were each assigned a random number using a computerized random number generator (Microsoft Ofce Excel 2003 Version 11.8; Microsoft Corporation, Redmond, WA) and had all patient identiers removed. The image sets were then randomly arranged by sorting the random numbers into ascending numerical order for viewing. The CT scans were obtained by using a Phillips Vitrea 2.0 CT scanner (Andover, MA) with 3-mm thick soft copy section thickness. The scanning protocol was use of a 16-section multidetector CT scanner (Brilliance 16 Power CT scanner; Philips Medical Systems, Cleveland, OH) with a detector width of 0.75 mm, pitch of 0.938, and rotation time of 0.5 seconds. The three-dimensional CT reconstructions and CTsimulated anteroposterior and Judet views of the pelvis were created by radiologists who used a standardized computer workstation and software (TeraRecon, San Mateo, CA). The reconstructions (image set C) and simulated views (image set D) allowed for multiple viewing directions as controlled by the viewer, as is typical clinically. To obtain the three-dimensional images, 2 3 2 mm thick images were used. All images were viewed with a General Electric Picture Archiving and Communication System (General Electric Corporation, Waukesha, WI) on an Agfa-Gevaert computerized workstation (AgfaGevaert Group, Mortsel, Belgium).

Image Evaluation
Each image set was evaluated in random order (as described previously) by four trauma fellowship-trained orthopaedic surgeons who routinely treat acetabular fractures in separate sessions. The diagnosis was recorded by each surgeon as either no fracture evident or as one of the 10 types of acetabular fractures of the Letournel and Judet classication system3: anterior wall, anterior column, posterior wall, posterior column, transverse, T-type, anterior column or
q 2010 Lippincott Williams & Wilkins

FIGURE 1. Three-dimensional computed tomography reconstructions of an acetabular fracture. Two views are shown (AB). The contralateral hemipelvis and femoral head have been removed for better visualization. The images can be rotated 360 in both the vertical and horizontal planes.

wall with posterior hemitransverse, both-column, posterior column and posterior wall, or transverse and posterior wall. Five weeks after all images had been reviewed by all surgeons, the gold standard diagnosis was determined by all four surgeons through a group consensus meeting. For the gold standard diagnosis, all 75 fractures were evaluated with all four imaging methods available; the viewers were still
www.jorthotrauma.com |

285

OToole et al

J Orthop Trauma  Volume 24, Number 5, May 2010

blinded to the patient identiers and to the diagnoses made by the individual surgeons. For any diagnosis with disagreement after discussion, a vote was taken to determine the gold standard diagnosis. Disagreement existed regarding only two of the 75 fractures.

Analysis
We rst compared the four imaging methods regarding ability to agree with the gold standard diagnosis. Considering that the frequencies of the 10 types of acetabular fractures are not equal in clinical practice nor were they in our data set, the analysis had to be weighted so that common fracture patterns did not falsely elevate the level of agreement. Agreement was assessed with the use of kappa statistics for multiple raters and nominal data. We next compared the four imaging methods regarding ability to consistently predict the diagnosis among surgeons. Again, agreement was assessed with the use of kappa statistics for multiple raters and nominal data. Additionally, a sensitivity analysis was conducted to ascertain that no single rater was driving the results of the study. Repetition of the analyses after exclusion of a single rater did not substantially change the overall study conclusions. Based on clinical experience of what fracture diagnoses are often debated by residents, fellows, and attendings, we hypothesized that three of the fracture types would be particularly likely to cause disagreement: associated both-column, T-type, and anterior column or wall with posterior hemitransverse. Therefore, we repeated these analyses after compressing the 10 fracture types down to eight. That is, we considered the three aforementioned fracture types as identical for purposes of this subanalysis to learn whether removing disagreement regarding the three fracture types would change our results.

Statistical Analysis
Agreement between the imaging methods and the gold standard diagnosis was evaluated with the use of kappa statistics for multiple raters and nominal data. The kappa values were interpreted by using the method described by Landis and Koch.14 The guidelines propose that kappa values of 0 to 0.20 indicate poor agreement, 0.21 to 0.40 fair agreement, 0.41 to 0.60 moderate agreement, 0.61 to 0.80 substantial agreement, and greater than 0.80 almost perfect agreement. Ninety-ve percent condence intervals were generated by using bootstrap techniques in a commercially available statistical package (STATA; SAS Institute, Inc., Cary, NC) yielding comparable results.

RESULTS
FIGURE 2. Simulated anteroposterior (A) and Judet views (BC) created from computed tomography (CT) scans of an acetabular fracture. The images simulate radiographs but are based on CT data. The images can be rotated 360 about vertical and horizontal axes. The images are not affected by body habitus, bowel gas, or the presence of contrast agent in the bladder. This is the same patient whose images are shown in Figure 3.

For each of the imaging methods (image sets A, B, C, and D), we assessed agreement with the gold standard diagnosis. Statistically signicant variation in agreement with the gold standard was shown based on imaging method (P , 0.01, analysis of variance). Image set B (Judet view plain radiographs alone) performed more poorly than did the other three image sets (P , 0.05, Duncans multiple range test) with only a 52% rate of agreement with the gold standard diagnosis (Table 1).
q 2010 Lippincott Williams & Wilkins

286

| www.jorthotrauma.com

J Orthop Trauma  Volume 24, Number 5, May 2010

Determining the Diagnosis of Acetabular Fractures

TABLE 1. Number of Raters Agreeing With Gold Standard Diagnosis by Imaging Method
Imaging Method A 4 of 4 3 of 4 2 of 4 1 of 4 0 of 4 Percentage agreement 30 17 17 7 4 71% B 19 13 16 9 18 52%* C 35 10 10 6 14 65% D 33 12 13 9 8 68%

TABLE 3. Chance-Adjusted Agreement on Diagnosis Among Imaging Methods*

Imaging Method A Kappa 95% condence interval 0.560 0.5190.630 B 0.512 0.4280.581 C 0.640 0.5840.689 D 0.659 0.6170.717

*Kappa statistics with 95% condence intervals. P , 0.05.

*Lower agreement with gold standard, P , 0.05.

The results reported in the previous paragraph are not satisfactory because they do not adjust for chance agreement. That is, raters are likely to agree regarding certain diagnoses (such as posterior wall fractures) that are highly prevalent, even if all the raters do is guess posterior wall for all the fractures. As discussed, we then adjusted for this effect and calculated kappa values, which account for the fact that agreement exists between raters based on chance alone (Table 2). When we used the more methodologically rigorous technique, image set B (Judet view plain radiographs alone) was again shown to be statistically signicantly less accurate than image sets A, C, and D (P , 0.05, analysis of variance, Duncans multiple range test) in terms of agreeing with the gold standard diagnosis. The image sets that included CT data (image sets A, C, and D) all had kappa values that indicated substantial agreement, whereas the image set that included plain radiographs alone (image set B) was rated as having moderate agreement. We next analyzed the number of times the raters agreed on a diagnosis (interobserver reliability) for each of the four imaging methods to determine which produced the most consistent results (Table 3). Image set B (plain radiographs alone) was statistically signicantly less consistent than the image sets that included CT data (image sets A, C, and D) (P , 0.05, analysis of variance, Duncans multiple range test). We repeated these analyses after compressing the classication system down to only eight fracture types by considering three of the fracture types to be the same (associated both-column, anterior column or wall with posterior hemitransverse, and T-type), as discussed previously. Doing so removed any differences from the data on these three fracture types that we thought might cause particularly high levels of

disagreement; however, the reanalysis yielded results similar to those of the initial analysis that analyzed all 10 fracture types (data not shown). We then analyzed the data by fracture type (Table 4); however, the large number of fracture types presents too small a sample size for statistical analysis regarding this issue.

DISCUSSION
The current standard radiographic assessment of a patient with an acetabular injury begins with three views of the pelvis: anteroposterior, obturatoroblique, and iliac oblique.47 With the addition of newer types of images such as axial view CT scans11,12,1520 and three-dimensional CT reconstructions,911 it was thought that not only would surgeons be able to better identify marginal impaction or loose bodies within the acetabulum but the diagnostic accuracy of the acetabular fracture patterns might improve.5,13 Little previous work has investigated the inuence of imaging techniques on the diagnosis of acetabular fractures. One study showed that reliability in classifying acetabular fractures by using the anteroposterior view pelvic radiograph alone was not improved with additional oblique (Judet) views21; however, the study participants were junior residents and community orthopaedic surgeons who presumably did not treat acetabular fractures, so the applicability of that study for acetabular surgeons is unclear. A study of orthopaedic trauma surgeons of various experience failed to show an advantage of axial view CT scans in improving the reliability of the classication system.5 That study did not analyze threedimensional CT-based data and used the surgeons opinion from the operating room as a gold standard. Two previous studies have investigated the role of threedimensional CT scans in the classication of acetabular fractures.13,22 In a smaller study of 20 fractures, only one of the ve participants was an orthopaedic trauma surgeon and the kappa values were low (kappa = 0.24), representing only fair agreement when using either plain radiographs or threedimensional CT scans.13 The kappa values were much lower than those previously reported in the literature5 and much lower than those in our study in which all the participants were orthopaedic trauma surgeons. A second larger study of 101 fractures evaluated two radiologists diagnoses and found results similar to those of our study with interobserver agreement rated as substantial (kappa = 0.70) with multidetector CT using three-dimensional reconstructions but only moderate (kappa = 0.42) with Judet lms alone.22 When
www.jorthotrauma.com |

TABLE 2. Chance-Adjusted Agreement With Gold Standard Diagnosis*

Imaging Method A Kappa 95% condence interval 0.647 0.6130.688 B 0.480 0.4370.546 C 0.620 0.5790.656 D 0.642 0.6070.709

*Kappa statistics with 95% condence intervals. Signicant difference at the P , 0.05 level.

q 2010 Lippincott Williams & Wilkins

287

OToole et al

J Orthop Trauma  Volume 24, Number 5, May 2010

TABLE 4. Accuracy in Diagnosis as a Function of Fracture Type

Fracture Type (by gold standard) Anterior wall Anterior column Posterior wall Posterior column Transverse T-type Anterior column or wall with posterior hemi-transverse Both-column Posterior column and posterior wall Transverse and posterior wall No. of Percent Percent Agreement Percent Percent Average Fractures Agreement Actual Judet Three-Dimensional Percent Agreement Across All Four (Total = 75) CT + Judet (A) Alone (B) CT (C) Simulated Judet (D) Imaging Modalities 0 5 20 2 4 3 7 11 4 19 75 89 63 81 33 43 86 63 68 45 79 63 75 0 14 89 50 36 75 85 50 100 17 29 95 75 46 65 95 63 81 58 39 89 75 47 65 87 60 84 27 31 90 66 49

CT, computed tomography; , not applicable.

comparing the diagnoses made during the study with the diagnoses from the operative reports, agreement was higher when based on CT scans than when based on radiographs, although the difference was statistically signicant for only one of the radiologists (P = 0.01 and 0.06). The applicability of the conclusions presented in these studies to diagnoses made by surgeons who regularly treat acetabular fractures is unknown. With our study, we retested the hypothesis that CT scans in comparison with plain radiographs alone will improve accuracy in classifying different acetabular fracture patterns. In contrast to ndings previously presented in the literature,5,13 we found that the diagnostic accuracy of plain radiographs alone was worse than that of CT scans, questioning the usefulness of plain radiographs for diagnosis of fracture pattern. Comparing the gold standard diagnosis with the four imaging sets, Judet view plain radiographs had a lower percentage of agreement (52% versus 71%, 65%, and 68%; P , 0.05, analysis of variance, Duncans multiple range test) and a worse kappa value than did the CT scans (P , 0.05, analysis of variance, Duncans multiple range test) (Tables 1 and 2). The kappa values for all three imaging methods that included CT scans showed substantial agreement, whereas the kappa values for the plain radiographs alone showed only fair agreement. Several possible explanations exist to explain why our ndings differ from those of previous work. In clinical practice, poor-quality radiographs are somewhat common, often because of body habitus, inadequate patient rotation, improper x-ray penetration, bowel gas, or contrast agent in the bladder (Fig. 3). We did not exclude any patients from our study because of poor lm quality; the included lms were those the surgeon actually used for management of the case. In that manner, we hoped to realistically characterize the inuence of the imaging studies on surgeons diagnosis. It is unclear whether previous studies included only patients with adequate plain lms, so it is possible that a selection bias was introduced in the study population of those other studies as a result of excluding patients with lower quality radiographs.

Another possible explanation for the difference between our ndings and the ndings of previous reports in the literature is the detail of the CT reconstructions. Modern software has improved such that recreated images have better resolution and more options are available for viewing the data. For example, the three-dimensional and simulated views can be rotated 360 in any direction. Also, the femoral head can be subtracted out of the acetabulum, providing the surgeon with an inside view of the pelvis unlike that of any plain radiograph. Other possibilities include differences in the details of our methodology. We used a statistical methodology that corrected for chance agreement and unequal fracture pattern frequencies, unlike previous studies. We used a consensus gold standard based on radiographs because we were not convinced that the operative surgeons opinion of the fracture morphology should be the gold standard, particularly for fractures with which the approach would not allow access to fracture lines on the opposite column. Our present study followed patients in a consecutive fashion, recreating clinical practice. This approach caused the number of certain types of acetabular fracture patterns to be seen more frequently, such as posterior wall and transverse posterior wall types (Table 4). Furthermore, a selection bias might have been introduced toward the radiologists obtaining and saving three-dimensional reconstructions of more interesting fracture patterns, because we had to exclude any patient who did not have three-dimensional reconstructions available for analysis, thus biasing the data set away from simple fractures and toward fractures that are more prone to disagreement regarding classication. However, we attempted to account for these factors by adjusting for chance agreement among imaging methods regarding diagnosis with our statistical methods. Strengths of our study include the randomized, blinded fashion in which four independent trauma fellowship-trained surgeons rated each image set for a diagnosis. We looked at a relatively large consecutive series of patients and included Judet view plain radiographs without regard for image quality, axial view CT scans, three-dimensional CT reconstructions, and CT-simulated anteroposterior and Judet views to best
q 2010 Lippincott Williams & Wilkins

288

| www.jorthotrauma.com

J Orthop Trauma  Volume 24, Number 5, May 2010

Determining the Diagnosis of Acetabular Fractures

FIGURE 3. Example of poor-quality obturatoroblique Judet view (A) and poor-quality iliac-oblique Judet view (B) plain radiographs obtained before operative treatment of an acetabular fracture. Image quality was limited by the patients body habitus and bowel gas. This is the same patient whose images are shown in Figure 2.

mimic images currently available to acetabular surgeons. We think our statistical methods adequately accounted for chance agreement and variation in the observed frequency of fracture patterns. Another component of evaluating imaging modalities is the cost. The professional fee plus the hospital charges for each of the image sets at our hospital in 2009 are as follows: image set A, $262; image set B, $687; and image sets C and D, $405. Many institutions obtain A, B, C, and D, which would cost $1354. These cost values represent only one hospitals fee schedule, and further work is required to conduct cost:benet analysis of these imaging modalities. The results of this study suggest future directions for research and clinical practice. With improving CT software
q 2010 Lippincott Williams & Wilkins

and different imaging modality techniques, the preoperative radiographic workup for patients with acetabular fractures might ultimately change. Perhaps after identifying an acetabular fracture on a routine anteroposterior view radiograph of the pelvis, the only other imaging modality required is CT. We have begun to study this protocol at our center. It has been our experience that it often is not practical or even possible to obtain additional radiographs of patients whose images are of poor quality, because the patients might be hemodynamically unstable or have associated injuries such as unstable spinal trauma. Furthermore, obtaining Judet view radiographs of awake patients typically is painful for the patients. Reshooting additional radiographs in pursuit of the perfect Judet view radiograph subjects patients to additional pain, radiation, increased costs, and preoperative evaluation time. CT scans can eliminate all the confounding factors and recreate a similar image virtually every time. One theoretical advantage of actual Judet lms is that after rotating the patient 45 onto a foam block to elevate one hip to obtain the radiographs, gravity provides stress to the hip that might reveal subluxation that might not be evident while the patient is supine; however, this phenomenon is unlikely to affect the diagnosis of fracture type and so will require further investigative research. The CT data obtained during the initial CT scanning can be used to construct traditional axial view images, threedimensional reconstructions, and simulated Judet views. The simulated Judet views have an appearance similar to that of a standard radiograph without the imperfections that a plain radiograph might have and can be used in the operating room or clinic (Fig. 2). The views can be digitally rotated at the surgeons discretion, recreating the perfect Judet view. Also, the femoral head can be digitally subtracted from threedimensional views, creating an internal acetabular view that no plain radiograph has been able to show (Fig. 1). It is important to note that our study investigated only the usefulness of these imaging modalities in making the diagnosis. Imaging is used to obtain other information such as the presence of joint impaction, intra-articular fragments, and the degree of fracture comminution, and to decide on the treatment and surgical approach, if surgery is indicated. Our study did not assess any of these additional issues. Further research is required to investigate whether three-dimensional reconstructions and CT-simulated views are reliable and clinically efcacious and whether they potentially limit radiation exposure, cost, and pain for patients who sustain acetabular fractures. ACKNOWLEDGMENTS We acknowledge the important assistance of Mary Zadnik Newell, OTR/L, Med, and Senior Editor and Writer, Dori Kelly, MA, Department of Orthopaedics, University of Maryland School of Medicine. REFERENCES
1. Letournel E. Fractures of the acetabulum: a study of a series of 75 cases [in French]. J Chir (Paris). 1961;82:4787. 2. Judet R, Judet J, Letournel E. Fractures of the acetabulum: classication and surgical approaches for open reduction: preliminary report. J Bone Joint Surg Am. 1964;46:16151646.

www.jorthotrauma.com |

289

OToole et al

J Orthop Trauma  Volume 24, Number 5, May 2010

three-dimensional computerized tomographic scan. J Orthop Surg (Hong Kong). 2000;8:3337. Landis JR, Koch CG. The measurement of observer agreement for categorical data. Biometrics. 1977;33:159174. Hayes CW, Balkissoon AR. Current concepts in the imaging of the pelvis and hip. Orthop Clin North Am. 1997;28:617642. Rai M, Firooznia H, Golimbu C, et al. The impact of CT in the clinical management of pelvic and acetabular fractures. Clin Orthop Relat Res. 1983;178:228235. Brandser E, Marsh JL. Acetabular fractures: easier classication with a systematic approach. AJR Am J Roentgenol. 1998;171:12171228. Hunter JC, Brandser EA, Tran KA. Pelvic and acetabular trauma. Radiol Clin North Am. 1997;35:559590. Martin JS, Marsh JL. Current classication of fractures: rational and utility. Radiol Clin North Am. 1997;35:491506. Resnik CS, Stackhouse DJ, Shanmuganathan K, et al. Diagnosis of pelvic fractures in patients with acute pelvic trauma: efcacy of plain radiographs. AJR Am J Roentgenol. 1992;158:109112. Petrisor BA, Bhandari M, Orr RD, et al. Improving reliability in the classication of fractures of the acetabulum. Arch Orthop Trauma Surg. 2003;123:228233. Ohashi K, El-Khoury GY, Abu-Zahra KW, et al. Interobserver agreement for Letournel acetabular fracture with multidetector CT: are standard Judet radiographs necessary? Radiology. 2006;241: 386391.

3. Letournel E. Acetabulum fractures: classication and management. Clin Orthop Relat Res. 1980;151:81106. 4. Letournel E, Judet R. Fractures of the Acetabulum, 2nd ed. New York: Springer; 1993. 5. Beaule PE, Dorey FJ, Matta JM. Letournel classication for acetabular fractures: assessment of interobserver and intraobserver reliability. J Bone Joint Surg Am. 2003;85:1704 1709. 6. Tornetta P III. Displaced acetabular fractures: indications for operative and nonoperative management. J Am Acad Orthop Surg. 2001;9:1828. 7. Mack LA, Harley JD, Winquist RA. CT of acetabular fractures: analysis of fracture patterns. AJR Am J Roentgenol. 1982;138:407412. 8. Mayo KA. Open reduction and internal xation of fractures of the acetabulum: results in 163 fractures. Clin Orthop Relat Res. 1994;305:3137. 9. Billet FP, Schmitt WG, Gay B. Computed tomography in traumatology with special regard to the advances of three-dimensional. Arch Orthop Trauma Surg. 1992;111:131137. 10. Burk DL Jr, Mears DC, Kennedy WH. Three-dimensional computed tomography of acetabular fractures. Radiology. 1985;155:183186. 11. Gautsch TL, Johnson EE, Seeger LL. True three dimensional stereographic display of 3D reconstructed CT scans of the pelvis and acetabulum. Clin Orthop Relat Res. 1994;305:138151. 12. Harley JD, Mack LA, Winquist RA. CT of acetabular fractures: comparison with conventional radiology. AJR Am J Roentgenol. 1982;138:413417. 13. Visutipol B, Chobtangsin P, Ketmalasiri B, et al. Evaluation of Letournel and Judet classication of acetabular fracture with plain radiographs and

14. 15. 16. 17. 18. 19. 20. 21. 22.

290

| www.jorthotrauma.com

q 2010 Lippincott Williams & Wilkins