Automatic quantification of scapular and glenoid morphology from CT scans using deep learning.

OBJECTIVES: To develop and validate an open-source deep learning model for automatically quantifying scapular and glenoid morphology using CT images of normal subjects and patients with glenohumeral osteoarthritis. MATERIALS AND METHODS: First, we used deep learning to segment the scapula from CT images and then to identify the location of 13 landmarks on the scapula, 9 of them to establish a coordinate system unaffected by osteoarthritis-related changes, and the remaining 4 landmarks on the glenoid cavity to determine the glenoid size and orientation in this scapular coordinate system. The glenoid version, glenoid inclination, critical shoulder angle, glenopolar angle, glenoid height, and glenoid width were subsequently measured in this coordinate system. A 5-fold cross-validation was performed to evaluate the performance of this approach on 60 normal/non-osteoarthritic and 56 pathological/osteoarthritic scapulae. RESULTS: The Dice similarity coefficient between manual and automatic scapular segmentations exceeded 0.97 in both normal and pathological cases. The average error in automatic scapular and glenoid landmark positioning ranged between 1 and 2.5 mm and was comparable between the automatic method and human raters. The automatic method provided acceptable estimates of glenoid version (R2 = 0.95), glenoid inclination (R2 = 0.93), critical shoulder angle (R2 = 0.95), glenopolar angle (R2 = 0.90), glenoid height (R2 = 0.88) and width (R2 = 0.94). However, a significant difference was found for glenoid inclination between manual and automatic measurements (p < 0.001). CONCLUSIONS: This open-source deep learning model enables the automatic quantification of scapular and glenoid morphology from CT scans of patients with glenohumeral osteoarthritis, with sufficient accuracy for clinical use.

Glenohumeral joint force prediction with deep learning.

Deep learning models (DLM) are efficient replacements for computationally intensive optimization techniques. Musculoskeletal models (MSM) typically involve resource-intensive optimization processes for determining joint and muscle forces. Consequently, DLM could predict MSM results and reduce computational costs. Within the total shoulder arthroplasty (TSA) domain, the glenohumeral joint force represents a critical MSM outcome as it can influence joint function, joint stability, and implant durability. Here, we aimed to employ deep learning techniques to predict both the magnitude and direction of the glenohumeral joint force. To achieve this, 959 virtual subjects were generated using the Markov-Chain Monte-Carlo method, providing patient-specific parameters from an existing clinical registry. A DLM was constructed to predict the glenohumeral joint force components within the scapula coordinate system for the generated subjects with a coefficient of determination of 0.97, 0.98, and 0.98 for the three components of the glenohumeral joint force. The corresponding mean absolute errors were 11.1, 12.2, and 15.0 N, which were about 2% of the maximum glenohumeral joint force. In conclusion, DLM maintains a comparable level of reliability in glenohumeral joint force estimation with MSM, while drastically reducing the computational costs.

Three-dimensional evaluation of the transverse rotator cuff muscle's resultant force angle in relation to scapulohumeral subluxation and glenoid vault morphology in nonpathological shoulders.

BACKGROUND: Static posterior subluxation of the humeral head (SPSH) results in glenohumeral osteoarthritis. Treatment strategies for SPSH with or without resulting osteoarthritis remain challenging. There is growing interest in evaluating the rotator cuff muscle volume, fatty infiltration, or forces in osteoarthritic shoulders with SPSH, mainly due to a possible transverse force imbalance. In nonpathological shoulders, the transverse angle of the rotator cuff muscle's resultant force may be associated with scapulohumeral alignment and glenoid vault morphology, despite an assumed transverse force balance. The purpose of this study was to assess the transverse rotator cuff muscle's resultant force angle (TRFA) and its relationship with the scapulohumeral subluxation index (SHSI) and selected glenoid vault parameters using computer modeling. METHODS: Computed tomography scans of 55 trauma patients (age 31 ± 13 years, 36 males) with nonpathological shoulders were analyzed and all measurements performed in 3-dimension. We placed landmarks manually to determine the humeral head center and the rotator cuff tendon footprints. The contours of the rotator cuff muscle cross-sectional areas were automatically predicted in a plane perpendicular to the scapula. Each rotator cuff muscle was divided into virtual vector fibers with homogeneous density. The resultant force vector direction for each muscle, corresponding to the rotator cuff action line, was calculated by vectorially summing the normalized fiber vectors for each muscle, weighted by the muscle trophic ratio. The resultant force vector was projected on the axial plane, and its angle with the mediolateral scapular axis was used to determine TRFA. The SHSI according to Walch, glenoid version angle (GVA), glenoid anteroposterior offset angle (GOA), glenoid depth, glenoid width, and glenoid radius were also evaluated. RESULTS: The mean values for TRFA, SHSI, GVA, GOA, glenoid depth, glenoid width, and glenoid radius were 7.4 ± 4.5°, 54.3 ± 4.8%, -4.1 ± 4.4°, 5.1 ± 10.8°, 3.3 ± 0.6 mm, 20 ± 2 mm, and 33.6 ± 4.6 mm, respectively. The TRFA correlated strongly with SHSI (R = 0.731, P < .001) and GVA (R = 0.716, P < .001) and moderately with GOA (R = 0.663, P < .001). The SHSI was strongly negatively correlated with GVA (R = -0.813, P < .001) and moderately with GOA (R = -0.552, P < .001). The GVA correlated strongly with GOA (R = 0.768, P < .001). In contrast, TRFA, SHSI, GVA, and GOA did not correlate with glenoid depth, width, or radius. CONCLUSION: Despite an assumed balance in the transverse volume of the rotator cuff muscles in nonpathological shoulders, variations exist regarding the transverse resultant force depending on the SHSI, GVA, and GOA. In healthy/nonosteoarthritic shoulders, an increased glenoid retroversion is associated with a decreased anterior glenoid offset.

Age- and sex-specific normative values of bone mineral density in the adult glenoid.

The objective of this study was to determine the normative bone mineral density (BMD) of cortical and trabecular bone regions in the adult glenoid and its dependence on the subject's age and sex. We analyzed computed tomography (CT) scans of 441 shoulders (310 males, 18-69 years) without any signs of glenohumeral joint pathology. Glenoid BMD was automatically quantified in six volumes of interest (VOIs): cortical bone (CO), subchondral cortical plate (SC), subchondral trabecular bone (ST), and three adjacent layers of trabecular bone (T1, T2, and T3). BMD was measured in Hounsfield unit (HU). We evaluated the association between glenoid BMD and sex and age with the Student's t test and Pearson's correlation coefficient (r), respectively. The lambda-mu-sigma method was used to determine age- and sex-specific normative values of glenoid BMD in cortical (CO and SC) and trabecular (ST, T1, T2, and T3) bone. Glenoid BMD was higher in males than females, in most age groups and most VOIs. Before 40 years old, the effect of age on BMD was very weak in both males and females. After 40 years old, BMD declined over time in all VOIs. This BMD decline with age was greater in females (cortical: r = -0.45, trabecular: r = -0.41) than in males (cortical: r = -0.30; trabecular: r = -0.32). These normative glenoid BMD values could prove clinically relevant in the diagnosis and management of patients with various shoulder disorders, in particular glenohumeral osteoarthritis and shoulder arthroplasty or shoulder instability, as well as in related research.