Abstract
Background
Preoperative planning software has been developed to measure glenoid version, glenoid inclination, and humeral head subluxation on computed tomography (CT) for shoulder arthroplasty. However, most studies analyzing the effect of glenoid positioning on outcome were done prior to the introduction of planning software. Thus, measurements obtained from the software can only be extrapolated to predict failure provided they are similar to classic measurements. The purpose of this study was to compare measurements obtained using classic manual measuring techniques and measurements generated from automated image analysis software.
Methods
Ninety-five two-dimensional computed tomography scans of shoulders with primary glenohumeral osteoarthritis were measured for version according to Friedman method, inclination according to Maurer method, and subluxation according to Walch method. DICOM files were loaded into an image analysis software (Blueprint, Wright Medical) and the output was compared with values obtained manually using a paired sample t-test.
Results
Average manual measurements included 13.8° version, 13.2° inclination, and 56.2% subluxation. Average image analysis software values included 17.4° version (3.5° difference, p < 0.0001), 9.2° inclination (3.9° difference, p < 0.001), and 74.2% for subluxation (18% difference, p < 0.0001).
Conclusions
Glenoid version and inclination values from the software and manual measurement on two-dimensional computed tomography were relatively similar, within approximately 4°. However, subluxation measurements differed by approximately 20%.
Keywords: glenohumeral arthritis, shoulder arthroplasty, glenoid version, glenoid inclination, humeral head subluxation, Walch classification
Introduction
Preoperative templating prior to shoulder arthroplasty is important for surgeons to gain an understanding of the osseous morphologies of the glenoid and humerus. Glenoid loosening remains a common reason for failure and reoperation after anatomic total shoulder arthroplasty.1–4 The accurate assessment of glenoid morphology prior to shoulder arthroplasty is critical in order to maximize implant longevity by decreasing the likelihood of implant malposition or failing to recognize significant humeral head subluxation. Traditionally, glenoid retroversion, glenoid inclination, and humeral head subluxation have been assessed manually on a two-dimensional computed tomography (CT) scan. These measurements were used to develop the Walch classification.5 Similarly, prognostic indices being associated with a greater risk of glenoid loosening, such as excessive retroversion or static posterior subluxation were also based on 2D CT scans.6
Several studies have demonstrated that correction of CT cuts to the scapular plane may lead to different values of these measurements,7–10 with one study reporting differences after as little as 1° of malrotation off the scapular axis.7 The use of automated 3D templating software is gaining popularity, particularly as patient-specific software becomes more sophisticated. Given the rising interest in automated software for preoperative templating, studies are emerging that have focused on the potential issue of accuracy. Chalmers et al.11 compared measurements between CT scans that were not corrected in the scapular axis, those that were corrected, and measurements delivered by one automated software program. The authors found that the software produced 3.5° of additional retroversion and significantly greater subluxation versus both the corrected and uncorrected values. However, that study focused solely on Walch type B2 glenoids. Boileau et al.12 reported very small differences (between 1° and 3°) when comparing manual measurements on two-dimensional and three-dimensional scans with an automated image analysis software; however, they did not report on subluxation measurements.
The purpose of this study was to compare measurements of version, inclination, and subluxation obtained using classic manual measuring techniques in comparison with an automated image analysis software in a large number of computed tomographies of shoulders with primary glenohumeral osteoarthritis encompassing the whole spectrum of the Walch classification.
Methods
Ninety-five consecutive shoulders with the diagnosis of primary glenohumeral osteoarthritis that underwent arthroplasty at a single institution between November 2013 and November 2016 were selected for the study (Mayo Clinic IRB#: 16-008350). All of the CT scans used in this study had been obtained at our institution using a standardized reformatting protocol that was developed in conjunction with our Radiology Department, with whom our senior surgeons work closely. The CT scan protocol used at our institution reformats the DICOM files in reference to the plane of the face of the glenoid. Outside CT scans were often of questionable quality with unknown methods of reformatting, and so were neither used for preoperative planning, nor for data analysis in this study. This ensured that the methodology within this study was standardized. CT scans from shoulders with a history of prior surgery or fracture that might have resulted in alterations of bone anatomy other than those secondary to glenohumeral osteoarthritis were excluded. The study was designed so that the sample would include all types of glenoid morphology considered in the Walch classification.5
These CT scans used in the study corresponded with 49 males and 46 females, with an average age at the time of their shoulder arthroplasty of 69 ± 9 years. An anatomic total shoulder arthroplasty had been performed in 64 (68%) shoulders, and a reverse shoulder arthroplasty in the remaining 31 (32%) shoulders. According to the modified Walch classification, there were 18 A1, 30 A2, 9 B1, 15 B2, 21 B3, 1 C, and 1 D glenoid.13
Manual measurements
The method of manual measurement was intended to mimic the manner in which it would be done in clinical practice (i.e., by a single observer familiar with the proper technique). The protocol for measurements was refined by the senior author and two shoulder and elbow fellows. Following several training sessions, an orthopedic surgery resident performed all of the manual measurements. Measurements were made using a proprietary image display program that allows tracing lines and measuring angles on two-dimensional images.
Glenoid version was measured on axial CT images according to the method described by Friedman et al.14 (Figure 1(a)). Sequential axial cuts were visualized from proximal to distal until the last cut in which the coracoid tip was visible was identified; the fourth cut distal to this coracoid reference cut was selected for measurements. A line was then traced connecting the anterior and posterior margins of the glenoid (“glenoid face horizontal line”), and the center point of this line was selected. A second line was traced connecting the center of the os trigonum with the center point of the glenoid face line (“scapular line”). A line of “neutral version” was traced perpendicular to the scapular line and through the anterior margin of the glenoid. The angle between the glenoid face horizontal line and the neutral version line was measured as the angle of glenoid version.
Figure 1.
Lines and angles used for measurements of glenoid version (a), glenoid inclination (b), and humeral head subluxation (c).
Glenoid inclination was measured on coronal CT images according to the method described by Maurer et al.15,16 (Figure 1(b)). The coronal cut that best captured the floor of the supraspinatus fossa was selected for measurements. A line was then traced connecting the superior and inferior margins of the glenoid (“glenoid face vertical line”). A second line was traced along the floor of the supraspinatus fossa (“supraspinatus fossa line”). The angle subtended between these two lines was measured as the angle of glenoid inclination.
Humeral head subluxation was measured according to the method of Walch et al.5 using the same cut selected for measurements of glenoid version (Figure 1c). A line was traced perpendicular to the “glenoid face horizontal line” through the center of this line (“glenoid axis line”). A second line was traced perpendicular to the glenoid axis line and at the junction between the medial third and the middle third of the humeral head (“articular cartilage base line”). The length of the portion of the articular cartilage base line behind the intersection with the glenoid axis line was measured as distance B, and divided by the length of the articular cartilage base line (distance A), obtaining a percentage of subluxation (B/A %).
Automated three-dimensional image analysis software measurements
The raw DICOM images of the same 95 CT scans used for manual measurements were loaded into an automated three-dimensional image analysis software (Blueprint, Wright Medical, Memphis, TN). Details regarding automated segmentation and measurements using this particular software have been described by Boileau et al.12 Briefly, this software uses a cloud of points to define the best fitting plane of the scapula, best-fitting sphere of the glenoid surface (average of neoglenoid and paleoglenoid for B2 glenoids), and best-fitting sphere of the humeral head. Glenoid version is measured as the angle between the scapular plane and the glenoid centerline projected on the transverse scapular plane. Glenoid inclination is measured in reference to a transverse line that runs through the supraspinatus fossa between the os trigonum and the middle of the glenoid vault. Humeral head subluxation is measured as the volume of humeral head sphere posterior to the midcoronal plane of the glenoid sphere in reference to the volume of the whole humeral head. Values of version, inclination, and subluxation automatically provided by this software for each of the 95 CT scans were recorded.
Statistical analysis
Each manually measured parameter was compared to the respective value generated by the automated three-dimensional image analysis software in a pairwise fashion using a paired sample t-test. The level of significance was established at p values < 0.05. Prior studies on the same topic showed statistical significance with samples of 31 shoulders11 and 60 shoulders,12 respectively. Our sample of 95 shoulders was expected to provide sufficient power.
Results
The average values provided by manual measurements in comparison with those obtained with the automated three-dimensional image analysis software are summarized in Table 1. Average values measured manually included 13.8° for version, 13.2° for inclination, and 56.2% for subluxation. Average values provided by the image analysis software included 17.4° for version (3.5° difference, p < 0.0001), 9.2° for inclination (3.9° difference, p < 0.001), and 74.2% for subluxation (17.9% difference, p < 0.0001).
Table 1.
Average and range of measurements for all 95 shoulders included in the study.
| Manual 2D | Automated 3D | Δ | p Value | |
|---|---|---|---|---|
| Glenoid version (°) | 13.8 ± 7.7 (0.1 to 39.3) | 17.4 ± 9.5 (1 to 46) | 3.5 ± 6.5 (0.1 to 20.1) | <0.0001 |
| Glenoid inclination (°) | 13.2 ± 10.1 ( − 13.7 to 44.6) | 9.2 ± 6.5 (0 to 40) | 3.9 ± 10.2 (0.2 to 37.6) | <0.001 |
| Humeral head subluxation (%) | 56.2 ± 7.0 (42.1 to 75.59) | 74.2 ± 13.0 (42 to 97) | 17.9 ± 11.1 (0.3 to 45.8) | <0.0001 |
Note: All values are reported as mean ± standard deviation (range).
Comparisons of values were further subanalyzed for the major categories of the Walch classification (Table 2). In addition, values were analyzed for the group of shoulders that had received an anatomic total shoulder arthroplasty and those that had received a reverse shoulder arthroplasty. For the anatomic total shoulder arthroplasty group, mean values for version as measured manually versus automatically were 12.5° versus 15.2°, 14.2° versus 9.0° for inclination, and 56.1% versus 71.7% for subluxation. For the reverse shoulder arthroplasty group, mean values for version as measured manually versus automatically were 16.7° versus 21.9° for version, 11.1° versus 9.7 for inclination, and 56.5% versus 79.1% for subluxation.
Table 2.
Average and range of measurements for each major Walch category and according to surgery performed.
| Walch A (n = 48) |
Walch B (n = 45) |
TSA (n = 64) |
RSA (n = 31) |
|||||
|---|---|---|---|---|---|---|---|---|
| M 2D | A 3D | M 2D | A 3D | M 2D | A 3D | M 2D | A 3D | |
| Glenoid version (°) | 9.8 ± 5.2 (0.1 to 21.7) | 13.2 ± 7.7 (1 to 33) | 17.9 ± 7.4 (3.9 to 39.3) | 22.0 ± 9.2 (2 to 46) | 12.5 ± 6.8 (0.1 to 32.3) | 15.2 ± 7.4 (1 to 32) | 16.7 ± 8.8 (0.8 to 39.3) | 21.9 ± 11.6 (2 to 46) |
| Glenoid inclination (°) | 11.0 ± 9.6 ( − 13.7 to 41.4) | 8.9 ± 5.2 (1 to 21) | 15.3 ± 10.5 (0.5 to 44.6) | 9.7 ± 7.8 (0 to 40) | 14.2 ± 9.6 ( − 10.1 to 41.4) | 9.0 ± 5.6 (1 to 32) | 11.1 ± 11.1 ( − 13.7 to 44.6) | 9.7 ± 8.2 (0 to 40) |
| Humeral head subluxation (%) | 53.1 ± 5.5 (42.1 to 63.9) | 67.2 ± 11.8 (42 to 92) | 59.9 ± 6.5 (46.64 to 75.6) | 81.4 ± 10.0 (52 to 97) | 56.1 ± 7.6 (42.1 to 75.6) | 71.7 ± 12.5 (42 to 95) | 56.5 ± 5.6 (46.6 to 69.3) | 79.1 ± 12.8 (48 to 97) |
Note: All values are reported as mean ± standard deviation (range). Walch class C and D were not displayed due to small numbers.
Discussion
Accurate preoperative measurements of glenoid version, glenoid inclination, and humeral head subluxation are important in the decision-making process regarding the relative indications of anatomic and reverse arthroplasty for glenohumeral osteoarthritis, as well as to plan for component position at the time of surgery. Classically, these parameters have been measured on two-dimensional CT scan slices. Currently, many surgeons have adopted software programs for preoperative planning that will measure version, inclination, and subluxation. The software selected for this study is particularly attractive because of its ability to segment the scapula and humerus automatically. However, in order to apply the measurements provided by these software packages to the body of literature published prior to the advent of preoperative planning software, it is important to understand how the values provided by these software packages compares with classic measurements.
The results of our study seem to indicate that measurements of glenoid version and inclination differ on average 4° when comparing manual measurements on two-dimensional CT scans and the values automatically provided by the specific software program utilized in our study. On the contrary, the average difference in humeral head subluxation was 18%, with the automated image analysis program providing larger subluxation values than those obtained through two-dimensional measurements according to the method described by Walch at the time of publication of the original Walch classification.5
The parameters measured in our study have been shown to be important in the field of shoulder arthroplasty. Laboratory studies have demonstrated increased strain, slipping, and lift-off when glenoid components are placed in excessive retroversion (over 15°).9,17,18 In clinical studies, glenoid loosening after anatomic total shoulder arthroplasty has been clinically correlated with excessive retroversion,4,19 and more specifically glenoid type B.5,6 The increased availability of CT imaging led to the development of commonly used classification systems such as those by Friedman et al.14 and Walch et al.5 Following this, several studies demonstrated that scapular orientation and morphology were essential considerations for glenoid evaluation.7–9,15,20,21 This led to an increasing interest in the use of three-dimensional CT reconstructed images to theoretically reduce this potential source of error.8,15,18,22,23
The use of 3D software based on CT images has been reported with promising results in an effort to automate, streamline and refine the templating process. As had occurred following the introduction of 2D and then 3D CT for preoperative assessment, studies have focused on the use of templating software, with particular focus on its applicability to patient-specific instrumentation.24 Ghafurian et al.25 effectively demonstrated that a customized computational framework was able to deliver accurate measurements of the glenoid parameters of interest with minimal inter-rater variability, though only on “normal” scapulae. The Cleveland Clinic group successfully applied a mathematical formula to predict a patient’s normal glenoid version based off of the anterior wall and plane of the scapula on the pathologic shoulder (i.e., contralateral side). Walch et al. reported promising results in regards to the precision and accuracy of guide-pin placement in vitro using patient-specific instrumentation created from the Glenosys software (Imascap, Brest, France).24 An in vivo report26 that evaluated the same software observed a mean error of 3.4° (SD = 5.1°) in the version of the glenoid component and 1.8° (SD = 5.3°) for inclination, with acknowledgment from the authors that these errors were higher than expected.
The automated image analysis software evaluated in this study has been previously assessed by others. Chalmers et al.11 evaluated glenoid version, inclination, depth, and humeral head subluxation values as measured on corrected and uncorrected CT’s, and compared the values also to those delivered by the Blueprint templating software. These authors only evaluated shoulders with a B glenoid pattern of deformity. They found that the software produced 3.5° higher retroversion versus the measurements obtained manually from the corrected and uncorrected CT images. Additionally, there was a 19° ± 8° significant difference between the software and uncorrected inclination, and higher subluxation values from the software versus both the uncorrected (13% ± 6%) and corrected measurements (11% ± 4%). More recently, Boileau et al.12 compared measurements on CT scans of 60 osteoarthritic shoulders using either a number of manual measurements or the same automated software used in our study. These authors investigated version and inclination, but they did not measure subluxation. Average version measurements where within 1.8–2.5° and average inclination measurements were within 0.2°. The results of our study are consistent with the results of Chalmers et al.11 and Boileau et al.12
Our study has several potential limitations. One might consider that relying on measurements from a single observer is a limitation. However, our goal was to mimic clinical practice as closely as possible, during which a single surgeon typically performs the measurements of version, inclination, and subluxation on his or her own. A second potential limitation of this study is that it used only one of the many methods reported to measure glenoid version or subluxation; however, we did select the methods that have been commonly reported in the literature. Our strengths include the large sample of shoulders, and the inclusion of all potential types of glenoid morphology according to the Walch classification.5
In conclusion, the automated image analysis software investigated in this study provided relatively similar values to those obtained through manual measurement on two-dimensional computed tomography for glenoid version and inclination, within approximately 4°. However, differences in measured subluxation percentage approximated 20%. This information will be of value when extrapolating information from publications published prior to the introduction of computer-based three-dimensional assessment of glenoid morphology and humeral head subluxation.
Figure 2.
Representative example of the values provided by the automated software analysis software investigated in this study (Blueprint, Wright Medical, Memphis, TN).
Declaration of Conflicting Interests
The author(s) declared the following potential conflicts of interest with respect to the research, authorship, and/or publication of this article: Dr Sanchez-Sotelo implant design Stryker Corporation, royalties. Dr Shukla is a paid consultant/speaker for Wright/Tornier. Other authors have nothing to disclose.
Ethical Review and Patient Consent
This study was approved by the Mayo Clinic Institutional Review Board (Mayo Clinic IRB#: 16-008350).
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.
References
- 1.Bulhoff M, Sattler P, Bruckner T, et al. Do patients return to sports and work after total shoulder replacement surgery? Am J Sports Med 2015; 43: 423–427. [DOI] [PubMed] [Google Scholar]
- 2.Chin PC, Hachadorian ME, Pulido PA, et al. Outcomes of anatomic shoulder arthroplasty in primary osteoarthritis in type B glenoids. J Shoulder Elbow Surg 2015; 24: 1888–1893. [DOI] [PubMed] [Google Scholar]
- 3.Raiss P, Bruckner T, Rickert M, et al. Longitudinal observational study of total shoulder replacements with cement: fifteen to twenty-year follow-up. J Bone Joint Surg Am 2014; 96: 198–205. [DOI] [PubMed] [Google Scholar]
- 4.Raiss P, Schmitt M, Bruckner T, et al. Results of cemented total shoulder replacement with a minimum follow-up of ten years. J Bone Joint Surg Am 2012; 94: e171. [DOI] [PubMed]
- 5.Walch G, Badet R, Boulahia A, et al. Morphologic study of the glenoid in primary glenohumeral osteoarthritis. J Arthroplasty 1999; 14: 756–760. [DOI] [PubMed] [Google Scholar]
- 6.Walch G, Young AA, Boileau P, et al. Patterns of loosening of polyethylene keeled glenoid components after shoulder arthroplasty for primary osteoarthritis: results of a multicenter study with more than five years of follow-up. J Bone Joint Surg Am 2012; 94: 145–150. [DOI] [PubMed] [Google Scholar]
- 7.Bryce CD, Davison AC, Lewis GS, et al. Two-dimensional glenoid version measurements vary with coronal and sagittal scapular rotation. J Bone Joint Surg Am 2010; 92: 692–699. [DOI] [PubMed] [Google Scholar]
- 8.Budge MD, Lewis GS, Schaefer E, et al. Comparison of standard two-dimensional and three-dimensional corrected glenoid version measurements. J Shoulder Elbow Surg 2011; 20: 577–583. [DOI] [PubMed] [Google Scholar]
- 9.Gross DJ, Golijanin P, Dumont GD, et al. The effect of sagittal rotation of the glenoid on axial glenoid width and glenoid version in computed tomography scan imaging. J Shoulder Elbow Surg 2016; 25: 61–68. [DOI] [PubMed] [Google Scholar]
- 10.Hoenecke HR, Jr, Hermida JC, Flores-Hernandez C, et al. Accuracy of CT-based measurements of glenoid version for total shoulder arthroplasty. J Shoulder Elbow Surg 2010; 19: 166–171. [DOI] [PubMed] [Google Scholar]
- 11.Chalmers PN, Salazar D, Chamberlain A, et al. Radiographic characterization of the B2 glenoid: the effect of computed tomographic axis orientation. J Shoulder Elbow Surg 2017; 26: 258–264. [DOI] [PubMed] [Google Scholar]
- 12.Boileau P, Cheval D, Gauci MO, et al. Automated three-dimensional measurement of glenoid version and inclination in arthritic shoulders. J Bone Joint Surg Am 2018; 100: 57–65. [DOI] [PubMed] [Google Scholar]
- 13.Bercik MJ, Kruse K, II, Yalizis M, et al. A modification to the Walch classification of the glenoid in primary glenohumeral osteoarthritis using three-dimensional imaging. J Shoulder Elbow Surg 2016; 25: 1601–1606. [DOI] [PubMed] [Google Scholar]
- 14.Friedman RJ, Hawthorne KB, Genez BM. The use of computerized tomography in the measurement of glenoid version. J Bone Joint Surg Am 1992; 74: 1032–1037. [PubMed] [Google Scholar]
- 15.Daggett M, Werner B, Gauci MO, et al. Comparison of glenoid inclination angle using different clinical imaging modalities. J Shoulder Elbow Surg 2016; 25: 180–185. [DOI] [PubMed] [Google Scholar]
- 16.Maurer A, Fucentese SF, Pfirrmann CW, et al. Assessment of glenoid inclination on routine clinical radiographs and computed tomography examinations of the shoulder. J Shoulder Elbow Surg 2012; 21: 1096–1103. [DOI] [PubMed] [Google Scholar]
- 17.Iannotti JP, Spencer EE, Winter U, et al. Prosthetic positioning in total shoulder arthroplasty. J Shoulder Elbow Surg 2005; 14(1 Suppl S): 111S–121S. [DOI] [PubMed] [Google Scholar]
- 18.Walch G, Mesiha M, Boileau P, et al. Three-dimensional assessment of the dimensions of the osteoarthritic glenoid. Bone Joint J 2013; 95B: 1377–1382. [DOI] [PubMed] [Google Scholar]
- 19.Ho JC, Sabesan VJ, Iannotti JP. Glenoid component retroversion is associated with osteolysis. J Bone Joint Surg Am 2013; 95: e82–e82. [DOI] [PubMed] [Google Scholar]
- 20.Bokor DJ, O’Sullivan MD, Hazan GJ. Variability of measurement of glenoid version on computed tomography scan. J Shoulder Elbow Surg 1999; 8: 595–598. [DOI] [PubMed] [Google Scholar]
- 21.Rouleau DM, Kidder JF, Pons-Villanueva J, et al. Glenoid version: how to measure it? Validity of different methods in two-dimensional computed tomography scans. J Shoulder Elbow Surg 2010; 19: 1230–1237. [DOI] [PubMed] [Google Scholar]
- 22.Sabesan VJ, Callanan M, Youderian A, et al. 3D CT assessment of the relationship between humeral head alignment and glenoid retroversion in glenohumeral osteoarthritis. J Bone Joint Surg Am 2014; 96: e64–e64. [DOI] [PubMed] [Google Scholar]
- 23.Terrier A, Ston J, Larrea X, et al. Measurements of three-dimensional glenoid erosion when planning the prosthetic replacement of osteoarthritic shoulders. Bone Joint J 2014; 96B: 513–518. [DOI] [PubMed] [Google Scholar]
- 24.Walch G, Vezeridis PS, Boileau P, et al. Three-dimensional planning and use of patient-specific guides improve glenoid component position: an in vitro study. J Shoulder Elbow Surg 2015; 24: 302–309. [DOI] [PubMed] [Google Scholar]
- 25.Ghafurian S, Galdi B, Bastian S, et al. Computerized 3D morphological analysis of glenoid orientation. J Orthop Res 2016; 34: 692–698. [DOI] [PubMed] [Google Scholar]
- 26.Gauci MO, Boileau P, Baba M, et al. Patient-specific glenoid guides provide accuracy and reproducibility in total shoulder arthroplasty. Bone Joint J 2016; 98B: 1080–1085. [DOI] [PubMed] [Google Scholar]