Knowledge of the premorbid glenoid shape and the morphological changes the bone undergoes in patients with glenohumeral arthritis can improve surgical outcomes in total and reverse shoulder arthroplasty. Several studies have previously used scapular statistical shape models (SSMs) to predict premorbid glenoid shape and evaluate glenoid erosion properties. However, current literature suggests no studies have used scapular SSMs to examine the changes in glenoid surface area in patients with glenohumeral arthritis. Therefore, the purpose of this study was to compare the glenoid articular surface area between pathologic glenoid cavities from patients with glenohumeral arthritis and their predicted premorbid shape using a scapular SSM. Furthermore, this study compared pathologic glenoid surface area with that from virtually eroded glenoid models created without influence from internal bone remodelling activity and osteophyte formation. It was hypothesized that the pathologic glenoid cavities would exhibit the greatest glenoid surface area despite the eroded nature of the glenoid and the medialization, which in a vault shape, should logically result in less surface area. Computer tomography (CT) scans from 20 patients exhibiting type A2 glenoid erosion according to the Walch classification [Walch et al., 1999] were obtained. A scapular SSM was used to predict the premorbid glenoid shape for each scapula. The scapula and humerus from each patient were automatically segmented and exported as 3D object files along with the scapular SSM from a pre-operative planning software. Each scapula and a copy of its corresponding SSM were aligned using the coracoid, lateral edge of the acromion, inferior glenoid tubercule, scapular notch, and the trigonum spinae. Points were then digitized on both the pathologic humeral and glenoid surfaces and were used in an iterative closest point (ICP) algorithm in MATLAB (MathWorks, Natick, MA, USA) to align the humerus with the glenoid surface. A Boolean subtraction was then performed between the scapular SSM and the humerus to create a virtual erosion in the scapular SSM that matched the erosion orientation of the pathologic glenoid. This led to the development of three distinct glenoid models for each patient: premorbid, pathologic, and virtually eroded (Fig. 1). The glenoid surface area from each model was then determined using 3-Matic (Materialise, Leuven, Belgium). Figure 1. (A) Premorbid glenoid model, (B) pathologic glenoid model, and (C) virtually eroded glenoid model. The average glenoid surface area for the pathologic scapular models was 70% greater compared to the premorbid glenoid models (P < 0 .001). Furthermore, the surface area of the virtual glenoid erosions was 6.4% lower on average compared to the premorbid glenoid surface area (P=0.361). The larger surface area values observed in the pathologic glenoid cavities suggests that sufficient bone remodelling exists at the periphery of the glenoid bone in patients exhibiting A2 type glenohumeral arthritis. This is further supported by the large difference in glenoid surface area between the pathologic and virtually eroded glenoid cavities as the virtually eroded models only considered humeral anatomy when creating the erosion. For any figures or tables, please contact the authors directly

Common post-operative problems in shoulder arthroplasty such as glenoid loosening and joint instability can be reduced by improvements in glenoid design shape, material choice and fixation method [1]. Innovation in shoulder replacement is usually carried out by introducing incremental changes to functioning implants [2], possibly overlooking other successful design combinations. We propose an automated framework for parametric analysis of implant design in order to efficiently assess different possible glenoid configurations. Parametric variations of reference geometries of a glenoid implant were automatically generated in SolidWorks. The different implants were aligned and implanted with repeatability using Rhino. The glenoid-bone models were meshed in Abaqus, and boundary conditions and loading applied via a custom-made Python script. Finally, another MATLAB script integrated and automated the different steps, extracted and analysed the results. This study compared the influence of reference shape (keel vs. 2-pegged) and material on the von Mises stresses and tensile and compressive strains of glenoid components with bearing surface thickness and fixation feature width of 3, 4, 5 or 6 mm. A total of 96 different glenoid geometries were implanted into a bone cube (E = 300 MPa, ν = 0.3). Fixed boundary conditions were applied at the distal surface of the cube and a contact force of 1000 N was distributed between the central nodes on the bearing surface. The implants were assigned UHMWPE (E = 1 GPa, ν = 0.46), Vitamin E PE (E = 800 MPa, ν = 0.46), CFR-PEEK (E = 18 GPa, ν = 0.41) or PCU (E = 2 GPa, ν = 0.38) material properties and the bone-implant surface was tied (Figure 1). The von Mises stresses, compressive and tensile strains for the different models were extracted. The influence of design parameters in the mechanical environment of the implant could be assessed. In this particular example, the 95. th. percentile values of the tensile and compressive strains induced by modifications in reference shape could be evaluated for all the different geometries simultaneously in form of radar plots. 2-pegged geometries (green) consistently produced lower tensile and compressive strains than the keeled (blue) configurations (Figure 2). Vitamin E PE and PCU glenoids also produced lower maximum von Mises stresses values than CFR-PEEK and UHMWPE designs (Figure 3). The developed method allows for simple, direct, rapid and repeatable comparison of different design features, material choices or fixation methods by analysing how they influence the mechanical environment of the bone surrounding the implant. Such tool can provide invaluable insight in implant design optimisation by screening through multiple potential design modifications at an early design evaluation stage and highlighting the best performing combinations. Future work will introduce physiological bone geometries and loading, a wider variety of reference geometries and fixation features, and look at bone/interface strength and osteointegration predictions