Comparative Analysis of 3D MRI and CT in Assessing Static Bone Loss and Dynamic Bipolar Interaction in Glenohumeral Instability

Abstract

Background:

Accurate assessment of glenoid bone loss and morphological variations is crucial for determining optimal surgical care pathways for shoulder instability. While 2-dimensional (2D) and 3-dimensional (3D) computed tomography (CT) has been the gold standard for evaluating static bone quality, 3D magnetic resonance imaging (MRI) has recently been proven reliable for these static assessments but remains unvalidated for dynamic, advanced morphological variables.

Purpose/Hypothesis:

The purpose of this study was to compare the utility of 3D MRI and 3D CT in measuring advanced, dynamic morphological variables in glenohumeral instability. It was hypothesized that 3D MRI would be comparable with 3D CT for assessing both static bone loss and dynamic morphological parameters.

Study Design:

Cohort study (diagnosis); Level of evidence, 2

Methods:

A total of 21 patients who had glenohumeral instability were included (mean age, 31.2 ± 14.9 years; 42.9% female; mean body mass index, 27.2 ± 5.8 kg/m²). All participants underwent preoperative 2D MRI and 2D CT scans with 3D MRI and 3D CT reformats. Patients were stratified based on a 10% subcritical bone loss threshold to evaluate measurement sensitivity between 3D MRI and 3D CT. Static imaging variables such as glenoid version, inclination, and bone loss (both glenoid and humeral) were measured. Dynamic variables—which are fixed values but are interactive/dependent on static measurements for calculation, including Hill-Sachs occupancy, glenoid track zones, and distance to dislocation were subsequently quantified. Paired t tests and χ² tests were employed to compare imaging modalities.

Results:

All patients demonstrated a Hill-Sachs lesion, with 15 having evidence of glenoid bone loss. There were no significant differences between 3D MRI and 3D CT in measuring glenoid or humeral bone loss, distance to dislocation, occupancy ratios, or glenoid track zones. Dynamic variables were consistent across imaging modalities, even after stratification by the 10% subcritical bone loss threshold. Both modalities accurately identified on/off track lesions and peripheral-track zones. Inter- and intrarater reliability was good to excellent for all CT and MRI measurements.

Conclusion:

Overall, 3D MRI is a validated and reliable alternative to 3D CT for preoperative evaluation of static glenohumeral bone loss and dynamic morphological variables in shoulder instability, allowing clinicians to choose the modality that best fits their practice.

Glenohumeral instability, which encompasses the broad topic of subluxation and dislocation in multidirectional planes, is a common yet difficult orthopaedic problem, affecting 1% to 2% of the general population. ¹⁸ This condition is frequently associated with varying degrees of bone loss in the glenoid and humeral head, referred to as Bankart and Hill-Sachs lesions, respectively.^10,34 Cadaveric and clinical studies alike have demonstrated how the degree of glenoid bone loss directly affects the operative success rate.^1,17,18,22 For example, in their landmark study, Burkhart and De Beer ⁶ reported that >25% critical bone loss of the glenoid results in a recurrent instability rate of 86% after arthroscopic Bankart repair. Similarly, Shin et al ⁴¹ defined their optimal value for critical bone loss to be 17.3%, with those above this threshold having significantly worse postoperative shoulder functional scores and Single Assessment Numeric Evaluation scores. More recently, Dekker et al ⁹ reported that patients with any degree of glenoid bone loss had a 4 times greater odds of recurrent dislocation after arthroscopic stabilization. They also determined that glenoid bone loss >15%, with >5 months of symptoms, and age <20 years were primary risk factors for recurrent instability. Finally, Shaha et al ³⁹ introduced the concept of a subcritical bone loss threshold, demonstrating that highly active patients with >13.5% glenoid loss had worse patient-reported outcomes, even in those without recurrent instability events. Overall, it is clear that (1) bone loss is a crucial factor to consider when managing shoulder instability and (2) the threshold value for critical bone loss has continued to decrease over the past decade in the orthopaedic literature. However, one major limitation is that many of these studies lack generalizability, often using inconsistent imaging modalities and heterogeneous techniques for measuring bone loss.

Recently, the concept of bone loss has been extended to advanced glenohumeral morphological variables, such as the glenoid track (GT), to define “peripheral-track” and “central-track” lesions. This approach aids in quantifying glenohumeral bone loss and predicting recurrent instability. For some surgeons, the concept of “off-track” lesions, which present a high risk of failure with arthroscopic Bankart repair alone, has even replaced the “engaging” Hill-Sachs lesion concept proposed by Burkhart and De Beer. ⁶ Additionally, lesions have been categorized into peripheral-track and central-track types based on the location of the medial edge of the Hill-Sachs lesion to the medial edge of the GT. The “distance to dislocation” (DTD) concept has also gained popularity in shoulder instability management. DTD is measured as the difference between the GT and the Hill-Sachs interval (HSI), and studies have shown that treatment for “on-track” lesions should not be uniform. Therefore, these advanced glenohumeral morphological variables may serve as valuable tools for surgical decision making, particularly in determining the need for remplissage and/or bony augmentation for the glenoid.

To this point, it is critical to develop highly specific and sensitive methodology for preoperative quantification of glenohumeral bone loss and comprehensive evaluation of morphological variations in the glenohumeral joint.^26,30 Using the fundamental concept that the inferior aspect of the glenoid exhibits a consistent true-circle shape, the “circle method” was introduced by Sugaya et al ⁴⁴ to quantify glenoid bone loss using 3-dimensional (3D)−reconstructed computed tomography (CT) images by subtracting the missing portion of a circle placed on an en face sagittal view of the glenoid. Huijsmans et al ²¹ validated and reproduced this circle method using both magnetic resonance imaging (MRI) and CT, reporting good estimates for the degree of bone loss for both modalities. Similarly, Gyftopoulos et al ¹⁹ showed that MRI can be used to accurately measure glenoid bone loss with a small margin of error, similar to both 2-dimensional (2D) CT and 3D CT. Finally, Lander et al ²⁶ used the circle method for novel 3D MRI image reconstruction and found that this modality was equivalent to 3D CT for estimating the degree of glenohumeral bone loss in shoulder instability. While MRI is preferred for preoperative assessment of soft tissue damage, there is a growing body of evidence suggesting MRI may be used for quantifying the advanced glenohumeral morphological variables for diagnosis, risk stratification, algorithmic treatment decision making, and successful prediction of shoulder instability.^21,30,45

This study seeks to expand upon previous research by conducting a comprehensive comparative analysis of different imaging modalities. Specifically, the aim is to assess the effectiveness of 3D MRI reformats and 3D CT reformats in the evaluation of static and dynamic glenohumeral morphological variables including glenohumeral bone loss, glenoid version and inclination, GT (on-, off-, and near-) zone of occupancy of the Hill-Sachs defect (zones 1-4), and DTD. We hypothesize that there will be no significant differences between 3D MRI and 3D CT for bone loss or morphological variables.

Methods

Patient Selection

Institutional review board approval with an exempt status was obtained from the study institution because of the retrospective nature of the present study. No informed consent was required. A retrospective review was conducted between August 2020 and January 2023 of 79 eligible patients aged 14 to 65 years with recurrent glenohumeral instability dislocation without previous surgery. All patients were pooled from a fellowship-trained surgeons’ (B.C.L.) repository at a single institution using standard arthroscopic techniques. The following inclusion criteria were then applied: patients >14 years, without previous surgery, who underwent both preoperative MRI and CT scans as part of their clinical assessment and surgical planning. Patients received CT and MRI evaluation as available per the present institution’s radiology schedule. Each patient underwent a comprehensive evaluation utilizing 4 distinct imaging modalities: 2D CT scans, 2D MRI scans, 3D CT reformats, and 3D MRI reformats. The specifications of the CT and MRI machines and image sequences were identical to those utilized in the Lander study. ²⁶

In brief, CT scans were performed using a 64−multidetector row CT with helical imaging, incorporating a 20% adaptive statistical iterative reconstruction for dose reduction. Key parameters included a 20-mm detector coverage, 0.531:1 pitch, 10.62 table speed, 0.625-mm slice thickness, 18-cm field of view, 120-kVp tube voltage, 0.8-second rotation time, automatic tube current modulation, and noise index of 22. A 3D osseous reconstruction was completed via institutional radiologists using Hounsfield unit−based segmentation and Tera Recon software per institutional protocol. For MRI scans, a Siemens Skyra 3T MRI scanner was utilized with a standard protocol including axial, coronal, and sagittal fat-saturated turbo spin-echo sequences with parameters of a 1-slab sequence, 20 distance factor, 180-mm² field of view, 0.7-mm slice thickness, repetition time of 10.5 ms, echo time of 3.5 ms, 256 × 100 matrix, 12° flip angle, 300-Hz bandwidth, and voxel dimensions of 0.7 × 0.7 × 0.7 mm. A specialized technologist used Philips IntelliSpace Portal to manually segment the anatomy for 3D reformats per institutional protocol.

After exclusion, the final cohort consisted of 21 patients, and the following variables were collected for every patient included in the study: sociodemographic data (age, sex, body mass index [BMI], tobacco use), Elixhauser comorbidity index (American Society of Anesthesiologists [ASA] score), and surgical characteristics (laterality, mechanism of injury, and athletic information if applicable) (Table 1)

Table 1.

Baseline Sociodemographic and Injury Characteristics ^a

Characteristic	Value
Age, y	31.2 ± 14.9 (15-65)
Sex, M/F, %	57.1/42.9
ASA	1.88 ± 0.58 (1-3)
BMI, kg/m2	27.2 ± 5.8 (18.5-39.2)
Tobacco use, %	33.3
Laterality, R/L, %	76.2/23.8
Mechanism, %
Athletics	38.1
Trauma	33.3
Occupational	19.1
Atraumatic	9.5
Sport, n
Lacrosse	1
Football	1
Swimming	1
Weightlifting	1
Basketball	1
Wrestling	1
Boxing	1
Gymnastics	1

^aData are presented as mean ± SD (range) unless otherwise indicated. ASA, American Society of Anesthesiologists; BMI, body mass index; F, female; L, left; M, male; R, right.

Two- and 3-Dimensional Radiological Variables, Measurements, and Analysis

Overall, all 4 imaging modalities (2D MRI, 2D CT, 3D MRI, 3D CT) were completed and obtained for all patients. The analysis involved the examination of various cross-sectional imaging variables, including glenoid version, inclination, glenoid bone loss, and humeral bone loss. These variables served as the input data for parametric calculations, specifically the determination of the GT, Hill-Sachs zone of occupancy, and the DTD values for each patient. The image measurements were independently performed by 2 fully trained and skilled reviewers (an orthopaedic surgery resident [J.T.-K.] and a medical student [I.U.]) both trained in musculoskeletal imaging. For intrarater reliability, one of the reviewers (J.T.-K.) then repeated these measurements 6 months after completion of the initial measurements and was blinded to the previous results. To determine interrater reliability, the second reviewer independently conducted the measurements and calculations for the same set of observations, without access to the initial results obtained by the first reviewer. Both reviewers worked collaboratively to ensure consistency in the assessment process, thereby allowing for an evaluation of the degree of agreement between their independent ratings. In addition, a senior sports medicine orthopaedic surgeon (B.C.L.) performed independent measurements and analyses as a final layer of validation.

2D Measurements

The measurement of 2D variables were performed using manual segmentation rather than automated software to best match the clinical practice of a single observer orthopaedic surgeon (B.C.L.). All measurements were made via institutional picture archiving and communication system (PACS) software using 2D images obtained from both MRI and CT scans. For quantification of glenoid version, the method developed by Friedman et al ¹⁵ and Choi et al ⁷ was utilized for all patients. This involved systematically analyzing sequential axial cuts from proximal to distal to best identify the tip of the coracoid as a reference point. Once identified, all measurements were performed using axial images that best identified the true center of the glenoid fossa and the most medial border of the scapula, with the goal of being 0 to 4 slices distal to the coracoid process. Three lines were then drawn: line 1, from the tip of the medial border of the scapula through the center of the glenoid fossa; line 2, a line perpendicular to line 1, which traversed the anterior margin of the glenoid; and line 3, which was the transverse line across the glenoid face connecting the anterior and posterior margins of the glenoid. In cases where a lesion was present, the center of the glenoid was estimated based on the remaining intact anatomy, and the anterior and posterior margins were identified to approximate the native glenoid geometry, ensuring that measurements remained reproducible and unaffected by the lesion. The glenoid version was determined to be the measured angle difference between line 3 and a 90° angle made by the intersection of lines 1 and 2 (Figure 1). Retroversion was determined to be a negative angle, while anteversion was a positive angle.

Figure 1.

Glenoid version and inclination measurement techniques for 2-dimensional (2D) computed tomography (CT) and magnetic resonance imaging (MRI). Glenohumeral version was measured using standard techniques on axial 2D CT and 2D MRI: the angle between a line connecting the anterior and posterior rims of the glenoid fossa (red line) and a reference line drawn along the scapular axis with 90° perpendicular to this axis (orange lines). The angle between the red and orange lines represents the version. This measurement accounts for posterior shifts in the glenoid center caused by anterior bone loss, ensuring accurate assessment of version despite defects. Inclination was measured using standard techniques on the coronal 2D CT and 2D MRI (green lines).

Glenoid inclination was assessed using the method proposed by Maurer et al, ³³ whereby coronal CT/MRI images were utilized, which best visualized the floor of the supraspinatus fossa and the largest glenoid diameter. A line was drawn along the inferior margin of the supraspinatus fossa. A second line was drawn connecting the superior and inferior margins of the glenoid fossa. The inclination was calculated by subtracting 90° from the angle between these 2 lines. Positive angles were defined as having superior tilt and negative angles defined as inferior tilt (Figure 1).

Quantification of Static Glenohumeral Bone Loss

Measurements of glenoid defects were conducted using all image modalities (2D CT, 2D MRI, 3D CT, 3D MRI). First, a central line was drawn along the long axis of the glenoid from the supraglenoid tubercle to the inferior glenoid rim using a sagittal en face view. A circle of best fit was then matched to the posteroinferior border of the glenoid. The diameter of this circle (D) was automatically generated using institutional PACS imaging software. A second line (d), which represented the size of the glenoid bone loss (in mm), was drawn from the intact anterior margin of the glenoid to the most anterior border of the circle of best fit (Figure 2). The percentage of glenoid bone loss was calculated as d/D× 100 (Appendix Figure A1, equation 2).

Figure 2.

Best-fit circle methodology for assessment of glenoid bone loss 3-dimensional (3D) magnetic resonance imaging (MRI) and 3D computed tomography (CT) glenoid reformation. Glenoid measurement for 3D reconstruction of MRI and CT utilizes sagittal en face view of the glenoid. A circle was placed over the inferior portion of the glenoid using margins as a guide for the circle. The glenoid height (green line) is drawn from the supraglenoid tubercle to the medial aspect of the inferior base. The quantification of glenoid bone loss (d) (red line) and circle diameter (D) (blue line)was quantified using the glenoid height line marking the center of the circle as reference. Percentage glenoid bone loss was determined as glenoid bone loss size/circle diameter × 100%. GT was determined through the following formula: 0.83 circle diameter – glenoid bone loss size.

The superoinferior diameter of the humeral head was measured in the coronal plane based on the slice that demonstrated the widest humeral head defined by the limits of the subchondral bone/cartilage. ³⁶ The Hill-Sachs defects were quantified using the methodology defined by Rowe et al. ³⁵ Hill-Sachs deformity was quantified using both 2D images and 3D reconstructions. For 2D measurements, Hill-Sachs lesions were best visualized on axial images, centered about the superior aspect of the proximal humerus at or above the level of the coracoid process. A best-fit circle was placed around the cortical margins of the humeral head at the widest point of the defect/HSI (defined as the maximal length between the medial margin of the Hill-Sachs lesion to the medial margin footprint of the rotator cuff on axial imaging), which was quantified in millimeters. To quantify humeral bone loss on the 3D-reformatted images, a best-fit circle method previously validated by Lander et al ²⁶ for 2D and 3D comparison was utilized (Figure 3). These static measurements served as the basis for deriving the dynamic morphological variables.

Figure 3.

Technique guide for measurement of humeral head assessment. (A) Two-dimensional (2D) magnetic resonance imaging (MRI) and 2D computed tomography (CT) measurement of Hill-Sachs maximal defect on axial 2D CT recorded in mm using perfect circle and length of the Hill-Sachs defect on axial slices and the line from the superior articular to inferior articular surface was quantified in mm on coronal imaging. (B) Three-dimensional (3D) CT measurement of humeral defect percentage from the 3D reformats using sagittal projection with best-fit (purple) circle. Green line determines humeral head height with perpendicular blue line to measure the residual humeral head and thereby the bone defect (red line). Percentage bone loss was determined through the formula A–B)/A) × 100].

GT, DTD, and Zone of Hill-Sachs Occupancy Measurements

GT was evaluated for all patients using the methodology extensively validated by Yamamoto et al,^50,51 Kurokawa et al, ²⁵ and Lander et al. ²⁶ In addition, the DTD, an extension of the GT concept that uses exact distances between GT and HSI of “near-track lesions” rather than binary on-track/off-track designation to predict likelihood of arthroscopic failure, was quantified using the methods established by Li et al. ²⁹ For 3D imaging, DTD was measured using high-resolution CT or MRI reconstruction. The GT was assessed using an en face view of the glenoid, and the HSI was determined as the distance between the medial edge of the Hill-Sachs lesion and the lateral margin of the humeral articular surface. DTD was then calculated as the difference between GT width and HSI, aligning the imaging plane to the lesion’s long axis to ensure accurate measurements and avoid errors inherent to 2D imaging. Finally, the concept of subcritical peripheral bone loss proposed by Shaha et al ³⁹ was applied to calculate the zone of occupancy of the Hill-Sachs defect (zones 1-4) for on-track lesions using the methods described by Yamamoto et al ⁵¹ (Figure 4).

Figure 4.

Glenoid track was divided into 4 zones based on the percentage of Hill-Sachs occupancy: zone 1, <25%; zone 2, 25% to <50%; zone 3, 50% to <75%; and zone 4, ≥75%. This Hill-Sachs lesion in zone 4 was determined to be a “peripheral-track” lesion. An example of Hill-Sachs lesion is represented by the ovoid shape (H). (Also see Appendix Figure A1).

Statistical Analysis

Statistical analysis was carried out to compare the differences between the various imaging modalities. Paired t tests and χ² tests were used to assess the significance of differences where applicable for continuous and categorical variables, respectively. The significance of differences involved calculating the z scores for each pairwise comparison between imaging modalities, then determining P values using the cumulative distribution function of the standard normal distribution. Statistical significance was set at P < .05 a priori. Intra- and interobserver reliability were also analyzed using Koo and Li classifications (poor, <0.5; moderate, 0.5-0.749; good, 0.75-0.90; excellent, >0.9). ²⁴ No commercial, industry-specific, or proprietary names were used unless deemed exceptionally necessary for context.

Results

Baseline demographic and clinical characteristics of 21 patients with shoulder instability were collected and are presented in Table 1. Overall, the mean patient age was 31.2 ± 14.9, with a higher proportion of men (57.1%) than women (42.9%). For Elixhauser comorbidity variables, the mean ASA score was 1.9 ± 0.6, the mean BMI was 27.2 ± 5.8, and tobacco use was present in the minority of patients (33.3%). Regarding surgical and injury data, the majority (76.2%) of patients had right shoulder involvement, and the most common mechanism of injury contributing to shoulder instability included athletic activities (38.1%), trauma (33.3%), occupational factors (19.1%), and atraumatic (9.5%) (Table 1).

Among the 21 patients meeting inclusion criteria in the present study, all 21 (100%) demonstrated a Hill-Sachs lesion. Additionally, 71.4% of patients (n = 15) exhibited some degree of glenoid bone loss. Importantly, there were no significant differences observed among the various 2D imaging modalities (MRI, CT) in terms of mean glenoid version or inclination. Furthermore, there were no statistically significant differences detected across these modalities in assessing bone loss (Table 2).

Table 2.

Quantification of Overall Parametric Measurements Between Imaging Techniques ^a

	2D CT	2D MRI	3D CT	3D MRI	P
Version, deg	–2.14	–2.47			.93
Inclination, deg	4.57	4.29			.97
Glenoid defect size, mm	2.67	2.59	2.54	2.55	.99
Glenoid defect, %	9.82	9.60	9.50	10.58	.93
Humeral bone loss, mm	8.30	8.39	8.09	7.42	.95
Humeral defect, %	18.55	18.54	18.91	18.35	.99
Hill-Sachs occupancy ratio, %	49.82	48.30	48.60	50.39	.99
Distance to dislocation, mm	9.99	10.04	9.97	8.78	.91

^a2D, 2-dimensional; 3D, 3-dimensional; CT, computed tomography; MRI, magnetic resonance imaging.

After conducting post hoc analyses using paired t tests and Tukey Honestly Significant Difference analyses, no statistically significant differences were observed among any of the imaging modalities for the static or dynamic measurements (Table 2). In terms of 2D glenoid morphology, there were no significant differences between 2D CT and 2D MRI measurements for mean glenohumeral version (P = .93) or inclination (P = .97). For a comparison of 3D measurements in glenohumeral bone loss, there were no significant differences in mean glenoid defect size for 3D CT (2.54 mm) and 3D MRI (2.55 mm) (P = .99) or humeral bone loss (3D CT, 8.09 mm; 3D MRI, 7.42 mm; P = .95). For dynamic measurements, the Hill-Sachs occupancy ratio means for 3D imaging were 48.60% (3D CT), and 50.39% (3D MRI), which were not statistically significant (P = .99). Finally, a comparison of mean DTD values were not statistically significant (3D CT, 9.97 mm; 3D MRI, 8.78 mm) (P = .91) (Table 2).

To evaluate whether 3D MRI could detect static glenohumeral bone loss and related dynamic variables, the cohort was divided based on a recently suggested subcritical glenoid bone loss threshold of 10%.^4,26 Differences and standard errors of difference were computed with a 95% CI. Results showed no significant differences between 3D MRI and 3D CT for measuring static or dynamic glenohumeral variables in either the <10% or the ≥10% bone loss groups (Tables 3 and 4).

Table 3.

Comparison of Imaging Modalities for Static and Dynamic Glenohumeral Variables in the <10% Bone Loss Group ^a

Measurement	Difference	Standard Error of Difference	t Test Paired 95% CI, After ANOVA		P
Glenoid defect, %			Lower CI	Upper CI
3D CT vs 3D MRI	0.27	1.97	–4.98	5.51	.89
Humeral defect, %
3D CT vs 3D MRI	0.04	3.09	–8.17	8.24	.99
Version
2D CT vs 2D MRI	1.72	2.13	–3.95	7.39	.42
Inclination
2D CT vs 2D MRI	0.33	1.02	–2.37	3.03	.75
Distance to dislocation, mm
3D CT vs 3D MRI	0.27	1.97	–4.98	5.51	.89
Zone 1-4, mean
3D CT vs 3D MRI	0.24	0.34	–0.66	1.13	.48

^a2D, 2-dimensional; 3D, 3-dimensional; ANOVA, analysis of variance; CT, computed tomography; MRI, magnetic resonance imaging.

Table 4.

Comparison of Imaging Modalities for Static and Dynamic Glenohumeral Variables in the >10% Bone Loss Group ^a

Measurement	Difference	Standard Error of Difference	t Test 95% CI, After ANOVA		P
Glenoid defect, %			Lower CI	Upper CI
3D CT vs 3D MRI	0.54	2.58	–6.53	7.61	.83
Humeral defect, %
3D CT vs 3D MRI	3.04	6.91	–15.91	22.00	.66
Version
2D CT vs 2D MRI	2.36	2.97	–5.78	10.49	.43
Inclination
2D CT vs 2D MRI	1.45	1.58	–2.888	5.781	.36
Distance to dislocation, mm
3D CT vs 3D MRI	1.82	2.92	–6.18	9.83	.53
Zones 1-4, mean
3D CT vs 3D MRI	0.25	0.43	–0.93	1.42	.56

^a2D, 2-dimensional; 3D, 3-dimensional; ANOVA, analysis of variance; CT, computed tomography; MRI, magnetic resonance imaging.

Our interrater reliabilities were all good to excellent, except for glenoid defect size (in mm) for 2D MRI and glenoid inclination on 2D MRI, which were both graded as moderate (Table 5). Our intrarater reliabilities were all excellent. A retrospective power analysis for 2-sided paired t tests revealed that with the present cohort of 21 patients, the study was able to identify a large effect size (Cohen d = 0.7) with a power of 0.80 and a significance level of alpha = .05.

Table 5.

Reliability (Intra- and Interrater) Analysis for Measurement Variables ^a

	Interrater (Correlation Coefficient)	Intrarater (Correlation Coefficient)		Interrater (Correlation Coefficient)	Intrarater (Correlation Coefficient)
Glenoid defect, %			Glenoid defect, mm
3D CT	0.92	0.99	2D CT	0.80	0.98
3D MRI	0.90	0.99	2D MRI	0.72	0.98
Humeral defect, %			Humeral defect, mm
3D CT	0.89	0.99	2D CT	0.97	0.99
3D MRI	0.84	0.99	2D MRI	0.96	0.96
Glenoid version			Glenoid inclination
2D CT	0.96	0.98	2D CT	0.82	0.85
2D MRI	0.90	0.97	2D MRI	0.72	0.92
DTD			Zone of HSO
3D CT	0.94	0.99	3D CT	1.0	1.0
3D MRI	0.98	0.99	3D MRI	1.0	1.0

^a2D, 2-dimensional; 3D, 3-dimensional; CT, computed tomography; DTD, distance to dislocation; HSO, Hill-Sachs occupancy; MRI, magnetic resonance imaging.

Discussion

The most important finding of the present study is that there were no differences between 3D MRI and 3D CT for quantifying static glenohumeral bone loss measurements or dynamic glenohumeral morphological variables in evaluating glenohumeral instability patients. In addition, no significant differences were detected when patients were stratified into 10% subcritical bone loss threshold groups (group 1, <10% bone loss; group 2: ≥10% bone loss). This suggests that 3D MRI may be just as effective and reliable as 3D CT, even for widely varying degrees of glenohumeral bone loss. This finding is supported by several recent studies indicating that 3D imaging modalities (MRI and CT) can be used interchangeably for assessing glenohumeral bone loss and dynamic markers in appropriately selected patients. The present study expands upon several others, which found no statistically significant differences in quantifying static glenohumeral bone loss between 3D MRI and 3D CT modalities.^2,21,43,48

Recently, 3D imaging, through CT and MRI, has been suggested to provide significant advantages over traditional 2D imaging by offering increased accuracy and detailed understanding of the spatial relationships and geometry of the shoulder joint. Unlike 2D imaging, which is limited to a single plane, 3D reconstructions allow for comprehensive visualization of the glenoid, humeral head, and surrounding structures from multiple angles, providing a clearer representation of complex anatomic features such as glenoid bone loss and Hill-Sachs defects. This enhanced spatial understanding may allow for more precise assessment and surgical planning, improving decision making. In contrast, 2D imaging is inherently limited, as it restricts the anatomy into a single plane, which can lead to measurement errors and inaccuracies when evaluating the full extent and positioning of defects. While 2D imaging may not capture the same spatial benefits, it remains useful and the gold standard in providing an overview of osseus structures in multiple slices and is often faster and more accessible in clinical settings. Additionally, 2D imaging is less resource-intensive, making it more readily available and cost-effective, particularly in situations where high detail is not required. However, for more complex cases involving significant bone loss or deformities, 3D imaging offers superior accuracy and spatial clarity. Despite its advantages, 3D imaging requires specialized training and expertise to ensure accurate interpretation. Manipulating 3D models demands a deeper understanding of shoulder anatomy, and without sufficient experience, there is a risk of misinterpretation. Therefore, while 2D CT and MRI imaging provides valuable insights for general assessments of static bone loss, 3D reconstruction imaging may be more reliable for detailed preoperative planning and complex anatomic evaluations, with its overall effectiveness dependent on the clinician’s proficiency.

To this point, recent evidence has suggested that both CT and MR can be reliably used to assess glenohumeral bone loss. ⁵² Lee at al ²⁸ reported that high spatial resolution MRI enabled highly sensitive and accurate evaluation of the glenoid, suggesting that MRI can replace CT in most cases. Similarly, Sgroi et al ³⁸ demonstrated noninferiority of MRI when compared with CT in a reliability study for measuring bone loss. Thus, the shift toward MRI centers on the advantage of being able to accurately assess variable degrees of glenohumeral bone loss and morphology, while simultaneously evaluating the surrounding soft tissues. For many clinicians, this would eliminate the need for any additional cross-sectional imaging studies (eg, CT scans), while significantly reducing cost, radiation exposure, and time.^26,30

Importantly, the high agreement and reliability between 3D MRI and 3D CT suggests that clinicians can choose the modality that best suits their patient, considering factors such as scanner availability, patient preference, cost, and ionizing radiation exposure. For example, Thacher et al ⁴⁵ highlighted how 3D CT offers shorter scan time, higher quality image reconstructions, and improved topographic evaluation compared with 2D CT, with the obvious downside of higher radiation dosage. Further, the multiplanar reconstruction of 3D CT also allows for a better sagittal profile of the glenoid and visuospatial analysis of any Hill-Sachs defects compared with 2D CT. ²⁶ Otherwise, clinicians must rely on 2D CT slices that exactly capture the optimal view of both the glenoid fossa and the humeral head for proper quantification of defect size. This not only has the potential for inconsistent reporting but can lead to significant variation in glenohumeral measurements, which decreases interrater reliability and may potentially lead clinicians to pursue unnecessary surgical pathways based on algorithmic bone loss thresholds.

Regardless of imaging modality, there will always be the potential for bias and variability in the measurements of glenohumeral bone loss and morphology. ⁴⁸ For example, Lander et al ²⁶ highlighted that notable clinical differences in critical bone loss (<10%) may only represent a few millimeters in measurement variation. Therefore, the goal of orthopaedic surgeons should be to identify the most accurate and reliable modality that minimizes the risk of bias and variability for an appropriate assessment of glenohumeral morphology and soft tissue structures. This is in direct agreement with de Mello et al, ⁸ who reported that 3D MRI and 3D CT had high intermodality and interrater agreement (intraclass correlation coefficients >0.94) with a bias of ≤1 mm for all measurements between observers. Therefore, if most patients with shoulder instability already receive a 2D MRI for soft tissue evaluation, the addition of 3D MRI with osseous reconstruction (which adds approximately 4 minutes of scan time) would allow for accurate bone measurements that may be at minimum noninferior and/or equivalent to 3D CT.^{27,32,43,46,48}

Since Yamamoto et al ⁵⁰ first introduced the concept of GT for evaluation of bipolar bone loss, further research has sought to define a threshold value for critical glenohumeral bone loss and the associated concept of on-track and off-track lesions. It is well-established that off-track lesions are significantly associated with increased rates of instability and need for revision surgery after arthroscopic Bankart repair.^11,31,37,47 While the consensus threshold value for critical bone loss remains >25%, ⁴² more recent data have emphasized the concept of subcritical bone loss. Shaha et al ³⁹ and others have described this as bone loss between 13.5% and 20%, which accurately predicts worse quality-of-life metrics after Bankart repair, independent of any recurrent instability.^5,39,51 Yamamoto et al ⁵¹ applied this concept to Hill-Sachs lesions, defining subcritical threshold zones in the GT based on the percentage of the Hill-Sachs occupancy, and reported that patients with Hill-Sachs occupancy ≥75% (ie, peripheral-track lesions) had inferior patient-reported outcomes regardless of recurrent instability. Recently, Li et al ²⁹ introduced the DTD calculation to assess the proximity of on-track lesions becoming off-track lesions. They identified that a DTD <8 mm represented a critical threshold for accurately predicting higher rates of failure for on-track shoulders after primary arthroscopic Bankart repair. Similarly, Barrow et al ¹ utilized 2D MRI to highlight a subgroup of patients exhibiting low (but not off-track) DTD measurements, defining them as near-track shoulders. As DTD approached 0 mm in this cohort, the risk of failure after arthroscopic Bankart repair significantly increased, suggesting the need for distinct treatment considerations. While the present study did not track functional or patient-reported outcomes, there is ongoing work from the authorship group in future studies to utilize these dynamic morphological variables in predictive modeling of clinical failure after surgical intervention.

Another important concept in shoulder instability is glenoid version and inclination. These variables are significantly increased in patients who have both anterior and posterior shoulder instability.^{13,16,20,40,49} Many studies have validated these findings, using a combination of both CT and MRI to define normal inclination ranging from 0° to 10° of superior tilt and glenoid version at 2° to 7° of retroversion.^12,14 For the present study, the calculations for version and inclination (CT-based version, –2.1 ± 5.5; MRI-based version, –2.5 ± 5.5; CT-based inclination, 4.6 ± 2.7; MRI-based inclination, 4.3 ± 2.8) are in direct agreement with the literature value ranges. In addition, the present study demonstrated there was no significant difference in measuring glenoid version (P = .93) or inclination (P = .97) between CT and MRI.

While there are many additional calculations and measurements that can be used to quantify glenohumeral bone loss, clinicians must at least be facile with the core concepts of glenoid inclination and version, GT, HSI, zone of Hill-Sachs occupancy, DTD, and critical versus subcritical bone loss thresholds. Further, to appropriately diagnose shoulder instability and severity stratification for algorithmic decision making, there needs to be a consensus on appropriate imaging modality for quantifying glenohumeral bone loss metrics and morphological variants. The present study suggests that 3D MRI is comparable and not significantly different compared with 3D CT for quantifying static and dynamic glenohumeral measurements. This is the first study of its kind to comprehensively evaluate the accuracy and reliability of multiple imaging modalities for measuring static bone loss and advanced dynamic morphological variables. The results of this study are consistent with the recent literature, which has popularized the utility of 3D MRI in shoulder instability. Finally, the variation in threshold values for critical and subcritical bone loss reported in the literature underscores the need for standardization in measurement techniques, an issue addressed in this study by using consistent methodology across imaging modalities and by stratifying patients based on the most recent subcritical bone loss thresholds in the literature (<10% and >10%).

Radiation Exposure From CT of the Shoulder

In an evaluation of radiation exposure data for CT scan, a consensus literature review revealed that the mean effective dose for a shoulder CT ranged from 2.1 to 10.8 mSV, which is equivalent to approximately 25.8 conventional chest radiographs. ³ The lifetime attributable risk of cancer development is 0.6/1000 for men and 0.73/1000 for women receiving a shoulder CT. ²³ Comparatively, shoulder MRI can perform high-resolution multiplanar imaging without the inherent risk of ionizing radiation. This makes MRI a safer option, particularly for patients requiring multiple imaging studies or those who are more vulnerable to the effects of radiation, such as children, pregnant patients, and other at-risk populations. Additionally, MRI provides superior soft tissue contrast, which can be crucial in glenohumeral instability assessment.

Limitations

There are several limitations of this study that need to be acknowledged. First, selection bias may exist due to the retrospective nature of the study and the inclusion of patients from a single-surgeon repository without a control or comparison cohort. Future prospective studies will focus on validating imaging acquisition and measurements between multiple centers. While efforts were made to standardize imaging protocols, variations in image acquisition could contribute to detection bias. Another limitation of our study is that the accuracy and reliability of 3D imaging are influenced by factors such as clinician proficiency, access to advanced imaging data and software with specially trained radiologists, and the quality and resolution of the scanners used, which can vary across clinical settings and affect the consistency of measurements. Finally, it is important to acknowledge the potential for beta error in comparisons that failed to detect statistically significant differences. A power analysis indicated that, with a cohort size of 21 patients, the study was able to identify a large effect size for this proof-of-concept study. However, the limited sample size remains a consideration in the interpretation of nonsignificant differences and clinically meaningful differences. Future research endeavors with larger, diverse cohorts and standardized methodologies will further advance the understanding of optimal imaging strategies in this clinical context.

Conclusion

This study suggests that 3D MRI is comparable in utility, precision, and sensitivity to 3D CT for evaluating static measurements and dynamic variables of glenohumeral instability. This provides clinicians with flexibility in choosing the most suitable modality based on practical considerations including cost, radiation exposure, patient preference, and physician comfort. These findings contribute unique and valuable insights to the ongoing discourse on imaging approaches in shoulder instability management.

Appendix Figure A1.

Equations for glenohumeral measurements.