Abstract
Background: Arthritic changes to glenoid morphology can be difficult to fully characterize on both plain radiographs and conventional two-dimensional computer tomography images. We tested the hypothesis that three-dimensional imaging of the shoulder would increase inter-rater agreement for assessing the extent and location of glenoid bone loss and also would improve surgical planning for total shoulder arthroplasty.
Methods: Four shoulder surgeons independently and retrospectively reviewed the preoperative computed tomography scans of twenty-four arthritic shoulders. The blinded images were evaluated with conventional two-dimensional imaging software and then later with novel three-dimensional imaging software. Measurements and preoperative judgments were made for each shoulder with use of each imaging modality and then were compared. The glenoid measurements were glenoid version and bone loss. The judgments were the zone of maximum glenoid bone loss, glenoid implant fit within the glenoid vault, and how to surgically address abnormal glenoid version and bone loss. Agreement between observers was evaluated with use of intraclass correlation coefficients and the weighted kappa coefficient (?), and we determined if surgical decisions changed with use of the three-dimensional data.
Results: The average glenoid version (and standard deviation) measured -17° ± 2.2° on the two-dimensional images and -19° ± 2.4° on the three-dimensional images (p < 0.05). The average posterior glenoid bone loss measured 9 ± 2.3 mm on the two-dimensional images and 7 ± 2 mm on the three-dimensional images (p < 0.05). The average anterior bone loss measured 1 mm on both the two-dimensional and the three-dimensional images. However, the intraclass correlation coefficients for anterior bone loss increased significantly with use of the three-dimensional data (from 0.36 to 0.70; p < 0.05). Observers were more likely to locate mid-anterior glenoid bone loss on the basis of the three-dimensional data (p < 0.05). The use of three-dimensional data provided greater agreement among observers with regard to the zone of glenoid bone loss, glenoid prosthetic fit, and surgical decision-making. Also, when the judgment of implant fit changed, observers more often determined that it would violate the vault walls on the basis of the three-dimensional data (p < 0.05).
Conclusions: The use of three-dimensional imaging can increase inter-rater agreement for the analysis of glenoid morphology and preoperative planning. Important considerations such as the extent and location of glenoid bone loss and the likelihood of implant fit were influenced by the three-dimensional data.
Clinical Relevance: We believe that these data support the concept that three-dimensional imaging techniques applied to the shoulder provide further information that may be useful to the surgeon during the planning of total shoulder arthroplasty.
Standard two-dimensional computed tomography studies have been shown to be helpful for the evaluation of the glenohumeral joint as well as for preoperative planning for total shoulder arthroplasty1-6. Three-dimensional computed tomography images have proved to be helpful for the interpretation of the complex osseous anatomy of the distal part of the humerus, the distal part of the radius, and elsewhere7-9.
Preoperative planning for total shoulder arthroplasty often focuses on the condition of the glenoid given that the ability to securely implant a prosthetic glenoid component depends on the remaining bone stock after glenoid preparation. Whether to anticipate the correction of pathologic glenoid version by reaming the "high side" or to augment extreme bone loss with a bone graft, for example, is typically influenced by a combination of experience, judgment, and preoperative imaging studies. As two-dimensional computed tomography images show only a portion of the anatomy in any given image, pathologic changes may be easier to evaluate with use of the global views provided by three-dimensional images.
The purpose of the present study was to evaluate if three-dimensional images of the shoulder would enhance preoperative surgical planning for total shoulder arthroplasty when compared with standard two-dimensional imaging. Our hypothesis was that, among four experienced shoulder surgeons, three-dimensional imaging would offer more consistent information such that the inter-rater agreement would increase both for characterizing arthritic glenoid pathology and for surgical planning.
Subjects
Forty-four consecutive patients who had undergone primary total shoulder arthroplasty for the treatment of end-stage glenohumeral arthritis between November 2005 and March 2007 at a shoulder reconstructive service at a large referral center were retrospectively reviewed. To meet the inclusion criteria of the present study, each patient was required to have end-stage glenohumeral arthritis of the shoulder and a preoperative computed tomography study as described below. The preoperative computed tomography studies were performed at the discretion of each surgeon as part of his or her normal preoperative evaluation to better evaluate glenoid bone volume. Patients with previous shoulder surgery, a history of glenohumeral trauma, or infection were excluded. Of the pool of forty-four consecutive patients, twenty-one did not have preoperative computed tomography studies. The remaining twenty-three nonconsecutive patients (twenty-four shoulders) met the inclusion criteria and were selected. The patients included seventeen men and six women with an average age of sixty-two years (range, forty-nine to seventy-eight years). Glenohumeral disease was secondary to osteoarthritis as described by Neer10 in nineteen shoulders, osteoarthritis secondary to glenoid hypoplasia in four, and rheumatoid arthritis in one. On the basis of the computed tomography data, there was a range of glenoid wear patterns, including no or minimal wear (three shoulders), eccentric posterior wear (thirteen), eccentric anterior wear (one), and both anterior and posterior wear (seven). All patient data were collected and analyzed with approval from and in accordance with our Internal Review Board.
Image Acquisition and Processing
A computed tomography scan that included the entire scapula was made for each shoulder. A sixty-four-slice computed tomography scanner was used for image acquisition (SOMATOM Sensation 64; Siemens Medical Solutions USA, Malvern, Pennsylvania). The patient was positioned supine on the computed tomography table. All scans were made with use of the same acquisition parameters (120 kV, 140 mA, 0.6-mm collimation, 512 × 512 matrix, no gantry tilt, and a 50-cm field of view). The field of view of each scan included the entire scapula at 1-mm increments in the axial plane11-13. The axial images were then used for three-dimensional post-processing.
Computed Tomography Analysis
The blinded computed tomography imaging data for each subject were presented individually to four experienced orthopaedic shoulder surgeons (J.J.S., M.J.C., J.J.B., and J.P.I.) for measurement and analysis. With regard to the experience level of the four observers, two had twenty-three years of experience each, performing more than 100 shoulder arthroplasties per year, and two had two and three years of experience, performing approximately forty shoulder arthroplasties per year. For the two-dimensional analysis, each computed tomography scan was loaded into commercially available imaging software (eFilm Workstation v2.1.2; Merge Healthcare, Milwaukee, Wisconsin) on a standard office computer (Pentium 4, 3 GHz). The software provides simple-to-use tools for making linear and angular measurements. This configuration represents the typical method with which computed tomography scans are currently viewed and interpreted at a standard computer workstation by a radiologist. For the three-dimensional analysis, the images were loaded into custom imaging software that we had developed12. This software renders each scan into a three-dimensional volumetric image that can be rotated and viewed from any angle. The software isolates the more radiodense osseous structures from the less dense soft tissues. All structures other than the scapula were therefore subtracted digitally, facilitating subsequent measurement and analysis. As is the case with the two-dimensional modality, simple-to-use tools for measurements are incorporated into the software. However, because the three-dimensional scapula can be manipulated in space as a free body, the plane of the scapula was defined by the observer with use of three points: the center of the glenoid fossa, the inferior pole of the scapular body, and the junction of the scapular spine at the medial border of the body5,12,14. Measurements in the three-dimensional modality were referenced from this subject-specific plane. The three-dimensional analysis was also done on a standard office computer (Pentium 4, 3 GHz). Each observer interpreted each blinded computed tomography study with use of the two-dimensional modality and then again, two weeks later, in random order, with use of the three-dimensional modality. Each observer made a series of measurements and judgments for each computed tomography study. A hands-on tutorial was provided to each observer for the use of the software, and each observer was allowed time to practice with the software before data were collected. Less than five minutes was needed to make the required measurements and judgments for each subject with use of the two-dimensional or three-dimensional modality. Two of the observers (J.J.S. and J.P.I.) were involved in the development of the three-dimensional software that was used in the present study, whereas the other two observers (M.J.C. and J.J.B.) were not.
Measurements
Glenoid version was assessed on the midglenoid slice (superior to inferior) with use of the technique described by Friedman et al.15. By convention, anteversion and retroversion were designated by positive and negative values, respectively. An assessment of both anterior and posterior maximum glenoid bone loss in the axial plane was recorded. To determine the extent of bone loss, the observer estimated where the joint line would have been prior to the glenoid erosions secondary to arthritis (Fig. 1, A). In some cases of isolated posterior bone loss, the observer might use the anterior glenoid margin as a guide to identify where the joint line had been. However, if the observer deemed both anterior and posterior bone loss (e.g., central erosion) to be present, an even greater reliance on subjective interpretation by the observer was required to determine the native joint line. In the axial plane, the maximum distances (to the nearest millimeter) from the arthritic joint line to the estimated original joint line for both the anterior and posterior halves of the glenoid were recorded. On the two-dimensional images, the measurements for each glenoid were therefore subject to the interpretation of each observer's estimation of glenoid wear as opposed to physiologic glenoid version. No presumption of physiologic version for any given glenoid was made. The differentiation between concentric posterior glenoid wear and native glenoid retroversion, for example, was left to the expertise of each observer. This judgment was meant to represent how a surgeon might commonly interpret a computed tomography study of glenohumeral arthritis and glenoid wear in the preoperative setting. However, the observers had additional tools at their disposal within the three-dimensional images that offered a means of estimating physiologic as opposed to pathologic glenoid version and, therefore, the extent of bone loss secondary to arthritis, as described below.
In order to describe where the greatest extent of anterior and posterior bone loss occurred, the glenoid face was divided into six zones: anterior and posterior halves as well as top, middle, and bottom thirds (Fig. 1, B). The location of maximum bone loss for both the anterior and posterior halves of the glenoid could be determined easily on the two-dimensional axial scan by noting in which third (superior to inferior) the axial image was located relative to the entire glenoid inferior-superior dimension.
Judgments
On the basis of their analysis of the computed tomography data, each observer was asked to select one of three surgical techniques that they would likely use to prepare the glenoid for the implantation of a glenoid component. In each case, the implanted glenoid component was required to be fully contained in the glenoid vault without perforations. The observer was not required to correct the glenoid to any predetermined amount of version but rather was asked to choose which of the given three surgical techniques they were likely to employ for that subject. Option 1 included reaming the glenoid to the subject's normal, physiologic glenoid version (e.g., by reaming parallel to the glenoid version in situ if version was deemed physiologic or by reaming the "high-side" to correct to physiologic version). This choice allowed the observer to accept a final position of glenoid retroversion if he or she deemed that retroversion was physiologic for that subject. It was up to the observer to determine what was physiologic for each shoulder. Therefore, if the observer determined that a particular glenoid was physiologically retroverted, for instance, the glenoid could be reamed in retroversion if desired. Option 2 included accepting a reamed position other than physiologic version for that subject if the observer believed that any additional reaming would compromise glenoid vault integrity or would medialize the joint line to an unacceptable extent. Option 3 included applying bone graft to the glenoid if the correction would be too great to achieve with reaming alone. Hemiarthroplasty is often performed for the treatment of shoulder arthritis; however, to facilitate comparisons and statistical analysis, it was not one of the available options in the present study. Each observer used and was familiar with the same total shoulder prosthetic system (Global Advantage; DePuy, Warsaw, Indiana) in his clinical practice. The polyethylene pegged glenoid component for this system (Anchor Peg Glenoid; DePuy) is available in six sizes, corresponding to the radius of curvature of the humeral head, and consists of a central large peg and three smaller collinear peripheral pegs. The central peg length measures 18 mm on the two smaller sizes and 21 mm on the four larger sizes. For all sizes, the collinear peripheral pegs are 7.5 mm long and have the same spacing. Each observer was asked if he believed that the pegged prosthetic glenoid implant of the chosen size would fit within the osseous vault without perforating the walls of the vault after preparation of the glenoid with the chosen method. Keeled glenoid components were not an available prosthetic choice in the present study.
Within the three-dimensional software, the observers had access to additional tools to help them interpret the computed tomography study. The three-dimensional shape of the glenoid vault endosteal volume, developed from the vaults of glenoids without arthritis, has been validated, both in cadaver modeling and in computer modeling similar to that in the present study, to be a highly conserved and consistent shape that can reliably predict both glenoid bone loss and physiologic glenoid version despite alterations of glenoid anatomy due to arthritis11,12,16. This vault shape was imported as a stereolithography free body and was placed into each glenoid by each observer. There were no automated, software-driven mechanisms for the placement of the vault model. Rather, each observer placed the vault model into each glenoid in a "best fit" manner according to the technique described previously11,12. The three-dimensional stereolithograph of the vault model was rotated in the sagittal plane so that its medial, tapered border aligned with the tapering dimensions of the medial aspect of the glenoid vault. In addition, by adjusting the version of the model in the axial plane, the anterior and posteromedial contours of the model were aligned with the adjacent endosteal borders of the osseous glenoid (Fig. 2). The reproducibility of the proper fit of the model within a glenoid vault by different observers and its subsequent ability to estimate bone loss due to arthritis and physiologic glenoid version despite the presence of arthritis has been validated previously11,12.
Given that each of the observers was familiar with the same total shoulder prosthetic system and used it in his clinical practice, a stereolithography model of the pegged glenoid implants (for each available size) was created and also was imported into the three-dimensional software. Like the vault model, the glenoid implant models could be manipulated in three dimensions, allowing an "implantation trial" of the component to assess for positioning and fit (Fig. 3).
All observers had training in the use of each imaging modality and were proficient in the use of each. Diagrams of glenoid version3,15 and bone loss measurements with descriptions were available during each evaluation session.
Statistical Analysis
Intraclass correlation coefficients were used to evaluate inter-rater agreement for continuous data. The kappa coefficient was used to evaluate inter-rater agreement for nominal data (zone of glenoid bone loss, glenoid implant fit, surgical decision). Kappa values (?) are useful for demonstrating inter-rater agreements and have been previously categorized as indicating slight agreement (0.00 to 0.20), fair agreement (0.21 to 0.40), moderate agreement (0.41 to 0.60), substantial agreement (0.61 to 0.80), and almost perfect agreement (=0.81)17. A value of 0 represents no agreement, and a value of 1.00 represents perfect agreement18,19. It should be noted that these guidelines regarding the level of agreement that might be considered acceptable are sensitive to the context in which the comparisons are posed. In general, decisions that are considered very important typically should be associated with a high level of agreement among experts and therefore are inherently subject to further interpretation.
The paired t test was used to compare differences in measured continuous data (glenoid version and extent of glenoid bone loss). Nonparametric tests were used when the data were not normally distributed. The level of significance was set at p < 0.05.
To evaluate whether the three-dimensional images significantly altered surgical decisions, the McNemar test of paired proportions was used. A two-tailed p value of <0.05 was considered significant.
A power analysis revealed that a minimum sample size of twenty-two shoulders evaluated by four surgeons (eighty-eight observations) would provide 95% power (a = 0.05, ß = 0.05) to detect a 10% difference in glenoid version and bone loss measurement.
Measurements
The computed tomography scans of twenty-four shoulders were reviewed independently by four observers (resulting in ninety-six observations) with use of the two-dimensional and then three-dimensional methods. Data are summarized in Table I.
Glenoid Version
The average glenoid version (and standard deviation) measured -17° ± 2.2° on the two-dimensional images and -19° ± 2.4° on the three-dimensional images (p < 0.05). Inter-rater agreement was very high for the measurement of glenoid version on two-dimensional and three-dimensional studies (0.95 and 0.96, respectively).
Magnitude of Glenoid Bone Loss
The average posterior glenoid bone loss measured 9 ± 2.3 mm on the two-dimensional images and 7 ± 2 mm on the three-dimensional images (p < 0.05). The average anterior glenoid bone loss measured 1 ± 0.7 mm on the two-dimensional images and 1 ± 0.7 mm on the three-dimensional images. For posterior bone loss, the intraclass correlation coefficient was not significantly different between the two-dimensional and three-dimensional images (0.83 and 0.80, respectively; p = 0.39). However, for anterior bone loss, the intraclass correlation coefficient increased significantly from 0.36 for the two-dimensional images to 0.70 for the three-dimensional images (p = 0.04).
Judgments
Location of Glenoid Bone Loss
With regard to the assessment of the zone of glenoid bone loss, there was slight agreement (?2D = 0.18) among observers with regard to anterior bone loss on the two-dimensional images; agreement increased to substantial (?3D = 0.74) on the three-dimensional images. With regard to the zone of posterior bone loss, agreement also improved from slight (?2D = 0.20) on the two-dimensional images to fair (?3D = 0.28) on the three-dimensional images. (?2D and ?3D are the kappa coefficients for the two-dimensional and three-dimensional images, respectively.)
With use of the three-dimensional data, the observers' judgments changed in nineteen (20%) of ninety-six cases when assessing the location of anterior bone loss and in seventeen (18%) of ninety-six cases when assessing the location of posterior bone loss. In cases in which observers' judgments about anterior bone loss changed, additional analysis revealed that the most common change (seen in fifteen of the nineteen cases) was from "no bone loss" on the two-dimensional images to "mid-anterior glenoid bone loss" (zone D) on the three-dimensional images; this change was significant (p = 0.02). For posterior bone loss, the most common change in judgment (seen in nine of the seventeen cases) was from "mid-posterior glenoid bone loss" (zone C) on the two-dimensional images to "no bone loss" on the three-dimensional images, but this change was not significant (p > 0.05).
Glenoid Prosthesis Fit
When asked if the prosthetic glenoid component would be fully contained by bone after version correction, inter-rater agreement increased from substantial (?2D = 0.67) on the two-dimensional images to almost perfect agreement (?3D = 0.87) on the three-dimensional images. In thirteen (14%) of ninety-six cases, the assessment of implant fit differed when the three-dimensional data were compared with the two-dimensional data. In twelve of those thirteen cases, the observer concluded the implant would not fit on the basis of the three-dimensional images; this finding was significant (p = 0.006).
Surgical Decision
Inter-rater agreement with regard to surgical decision-making increased from fair (?2D = 0.29) with use of the two-dimensional data to moderate (?3D = 0.49) with use of the three-dimensional data. Surgical decisions that were made on the basis of two-dimensional data differed from those that were made on the basis of three-dimensional data in thirty-seven (39%) of ninety-six cases. In cases in which the decisions based on two-dimensional and three-dimensional data differed, the most frequent change (seen in seventeen of thirty-seven cases) was from "accept a reamed position other than the physiologic version" on the two-dimensional images to "ream to physiologic glenoid version" on the three-dimensional images. If the responses differed from the two-dimensional images to the three-dimensional images and "bone graft the glenoid" was one of the responses, the decision to use a bone graft was based on the three-dimensional view in nine of ten instances. There was a unanimous decision to use a bone graft in the glenoid in five cases in which the two-dimensional images were used and in eight cases in which the three-dimensional images were used. Compared with the two less experienced surgeons, the two more experienced surgeons did not demonstrate significantly increased agreement with regard to surgical decision-making (? = 0.56 and ? = 0.61, respectively; p = 0.3).
In the present study, the use of three-dimensional imaging increased the inter-rater agreement for the analysis of glenoid morphology and preoperative planning. The magnitude of anterior glenoid bone loss, the location of both anterior and posterior maximum glenoid bone loss, the judgment of preoperative prosthetic glenoid implant fit, and the selection of surgical technique for glenoid preparation all demonstrated less observer variation when the three-dimensional images were used.
Although correlation of our data with actual surgical results was not a part of our methodology, we believe that the use of such three-dimensional imaging data is clinically relevant given that greater agreement in the measurements and judgments was demonstrated among four experienced shoulder surgeons. We suggest that the greater appreciation of the three-dimensional anatomy of the glenoid, for example, allowed the observers to adjust potential surgical techniques to help to avoid implant malposition or improper fit. Similarly, previous studies investigating the utility of three-dimensional imaging techniques have shown greater observer agreement in terms of the depiction of fracture patterns and the recommendation of fixation strategies7,8.
On the average, posterior glenoid bone loss appeared to be significantly greater on the two-dimensional images as compared with the three-dimensional images. This finding suggests that the observers were more likely to estimate a physiologic joint line at a more lateral level on the two-dimensional images. As in previous studies in which the glenoid vault model was used as a proxy to estimate normal glenoid anatomy11,12,16, we used the vault model as a template in association with the three-dimensional images to better approximate the position and orientation of the native joint line, which allowed for greater agreement in terms of preoperative planning. In contrast, the observers had no vault model template with which the native joint line could be estimated when using the two-dimensional images. Greater subjectivity during the two-dimensional image interpretation with regard to physiologic retroversion or concentric posterior glenoid bone loss was therefore more likely. For a given shoulder with both retroversion and arthritis, it is possible for an observer to interpret this combination as either being concentric posterior wear or being physiologically retroverted with arthritis (Fig. 2). We believe that the issue of interpretation precisely adds to variability in surgical decision-making and technique as the surgeon may choose to ream in situ at that degree of retroversion or to ream the high side to make the version closer to zero. With the two-dimensional images, the observers were left to speculate as to the true extent of glenoid bone loss. However, we suggest that with the current three-dimensional tools such as the vault model, physiologic glenoid retroversion was better differentiated from concentric posterior wear and the true extent of bone loss was determined with greater accuracy and reliability.
We did not assume that all glenoids "should be" at 0° of version or that all glenoids are normally slightly retroverted. The available literature suggests a wide array of physiologic versions are possible3,6,20-23. It is this variability that we believe may increase the variation in surgical decision-making among experts with regard to how much to ream. With the availability of the three-dimensional imaging and tools such as the vault model as a glenoid template, this variability was decreased significantly.
The three-dimensional imaging described in the present study is a result of processing already completed computed tomography images and can be done at any time after the original computed tomography scan has been performed. Therefore, there is no greater patient time in the scanner and no additional cost. Currently, the time from completion of a computed tomography scan to a three-dimensional study that is ready for analysis (including the time for downloading the images, compiling the separate computed tomography images into one volumetric image, and isolating the scapula and humerus to be ready for measurements) is about five minutes.
We did not propose a "gold standard" answer for any of the surgical decision choices. We therefore cannot conclude that some observers were correct or incorrect in their surgical plans. We believe that comparing the observers' choices to the actual surgical technique ultimately employed for each patient would be misleading and inappropriate given that the actual surgical decision also varies according to surgeon experience and judgment. Rather, we can state on the basis of these data that qualified observers gleaned additional information from the three-dimensional images, resulting in changes in measurements and surgical judgments and, in most cases, greater agreement.
A limitation of the present study is that the computed tomography scans were made at the discretion of the surgeon and may have been preferentially made in cases of more severe glenoid disease. Thus, the differences between the two-dimensional and three-dimensional imaging may be less apparent in cases of mild or absent glenoid erosions.
Two of the four observers were involved in the development of the three-dimensional software used in the present study and therefore have a potential conflict of interest, and we acknowledge the potential for observer bias in our results. However, we believe that our study design, which specifically incorporated an equal number of observers with expertise in shoulder arthroplasty, independent of this software's development and with no conflict of interest, effectively limited this potential observer bias.
Clearly, surgical decision-making in shoulder arthroplasty is greatly influenced by training, experience, and viewpoint. Adding further variability, opinions differ among experts as to the extent of glenoid reaming and when glenoid bone-grafting is appropriate. The potential benefits of the three-dimensional images may, however, not be as apparent in cases in which no or mild glenoid deformity exists. Furthermore, three-dimensional computed tomography images can now be routinely made with minimal additional expense and no additional radiation exposure7,8. Commercially available three-dimensional computed tomography software does not have all of the tools (e.g., the glenoid vault model or the prosthetic glenoid implant model) or the ability to duplicate the data or measurements made with use of the surgical simulator software used in the current study. Additional efforts to expand such tools into modern imaging software may prove beneficial beyond the simple appreciation of the three-dimensional anatomy. We believe that the present study supports the concept, and our personal observation, that three-dimensional imaging techniques applied to the shoulder provide additional information to the surgeon that may help to facilitate more consistent surgical planning of total shoulder arthroplasty. 
Note: The authors thank Michelle Secic, MS, for her assistance with the statistical analysis of the data.
Green A, Norris TR. Imaging techniques for glenohumeral arthritis and glenohumeral arthroplasty. Clin Orthop Relat Res.1994;307:7-17.3077
1994
[PubMed]
Iannotti JP, Norris TR. Influence of preoperative factors on outcome of shoulder arthroplasty for glenohumeral osteoarthritis. J Bone Joint Surg Am.2003;85:251-8.85251
2003
Nyffeler RW, Jost B, Pfirrmann CW, Gerber C. Measurement of glenoid version: conventional radiographs versus computed tomography scans. J Shoulder Elbow Surg.2003;12:493-6.12493
2003
[CrossRef]
Ogawa K, Yoshida A, Ikegami H. Osteoarthritis in shoulders with traumatic anterior instability: preoperative survey using radiography and computed tomography. J Shoulder Elbow Surg.2006;15:23-9.1523
2006
[CrossRef]
Welsch G, Mamisch TC, Kikinis R, Schmidt R, Lang P, Forst R, Fitz W. CT-based preoperative analysis of scapula morphology and glenohumeral joint geometry. Comput Aided Surg.2003;8:264-8.8264
2003
[CrossRef]
Couteau B, Mansat P, Mansat M, Darmana R, Egan J. In vivo characterization of glenoid with use of computed tomography. J Shoulder Elbow Surg.2001;10:116-22.10116
2001
[CrossRef]
Doornberg J, Lindenhovius A, Kloen P, van Dijk CN, Zurakowski D, Ring D. Two and three-dimensional computed tomography for the classification and management of distal humeral fractures. Evaluation of reliability and diagnostic accuracy. J Bone Joint Surg Am.2006;88:1795-801.881795
2006
[CrossRef]
Harness NG, Ring D, Zurakowski D, Harris GJ, Jupiter JB. The influence of three-dimensional computed tomography reconstructions on the characterization and treatment of distal radial fractures. J Bone Joint Surg Am.2006;88:1315-23.881315
2006
[CrossRef]
Newton PO, Hahn GW, Fricka KB, Wenger DR. Utility of three-dimensional and multiplanar reformatted computed tomography for evaluation of pediatric congenital spine abnormalities. Spine.2002;27:844-50.27844
2002
[CrossRef]
Neer CS 2nd. Replacement arthroplasty for glenohumeral osteoarthritis. J Bone Joint Surg Am.1974;56:1-13.561
1974
Scalise JJ, Bryan J, Polster J, Brems JJ, Iannotti JP. Quantitative analysis of glenoid bone loss in osteoarthritis using three-dimensional computed tomography scans. J Shoulder Elbow Surg.2008;17:328-35.17328
2008
[CrossRef]
Scalise JJ, Codsi MJ, Bryan J, Iannotti JP. The three-dimensional glenoid vault model can estimate normal glenoid version in osteoarthritis. J Shoulder Elbow Surg.2008;17:487-91.17487
2008
[CrossRef]
Bushberg JT, Seibert JA, Leidholdt EM Jr, Boone JM. The essential physics of medical imaging. 2nd ed. Philadelphia: Lippincott Williams and Wilkins; 2002. Tomographic reconstruction; p 346-57.
2002
Kwon YW, Powell KA, Yum JK, Brems JJ, Iannotti JP. Use of three-dimensional computed tomography for the analysis of the glenoid anatomy. J Shoulder Elbow Surg.2005;14:85-90.1485
2005
[CrossRef]
Friedman RJ, Hawthorne KB, Genez BM. The use of computerized tomography in the measurement of glenoid version. J Bone Joint Surg Am.1992;74:1032-7.741032
1992
Codsi MJ, Bennetts C, Gordiev K, Boeck DM, Kwon Y, Brems J, Powell K, Iannotti JP. Normal glenoid vault anatomy and validation of a novel glenoid implant shape. J Shoulder Elbow Surg.2008;17:471-8.17471
2008
[CrossRef]
Posner KL, Sampson PD, Caplan RA, Ward RJ, Cheney FW. Measuring interrater reliability among multiple raters: an example of methods for nominal data. Stat Med.1990;9:1103-15. Erratum in: Stat Med. 1992;11:1401.91103
1990
[CrossRef]
Landis JR, Koch GG. The measurement of observer agreement for categorical data. Biometrics.1977;33:159-74.33159
1977
[CrossRef]
Landis JR, Koch GG. An application of hierarchical kappa-type statistics in the assessment of majority agreement among multiple observers. Biometrics.1977;33:363-74.33363
1977
[CrossRef]
Bokor DJ, O'Sullivan MD, Hazan GJ. Variability of measurement of glenoid version on computed tomography scan. J Shoulder Elbow Surg.1999;8:595-8.8595
1999
[CrossRef]
Churchill RS, Brems JJ, Kotschi H. Glenoid size, inclination, and version: an anatomic study. J Shoulder Elbow Surg.2001;10:327-32.10327
2001
[CrossRef]
De Wilde LF, Berghs BM, Audenaert E, Sys G, Van Maele GO, Barbaix E. About the variability of the shape of the glenoid cavity. Surg Radiol Anat.2004;26:54-9.2654
2004
[CrossRef]
Randelli M, Gambrioli PL. Glenohumeral osteometry by computed tomography in normal and unstable shoulders. Clin Orthop Relat Res.1986;208:151-6.208151
1986