Computer Aided Tools for the Design and Planning of Personalized Shoulder Arthroplasty

Shoulder arthroplasty is the third most common joint replacement procedure, after knee and hip arthroplasty, and currently the most rapidly growing one in the orthopaedic field. The main surgical options include total shoulder arthroplasty (TSA), in which the normal joint anatomy is restored, and, for patients with a completely torn rotator cuff, reverse shoulder arthroplasty (RSA), in which the ball and the socket of the glenohumeral joint are switched. Despite the recent progress and advancement in design, the reported rates of complication for RSA are higher than those of conventional shoulder arthroplasty. A patient-specific approach, in which clinicians adapt the surgical management to patient characteristics and preoperative condition, e.g. through custom implants and pre-planning, can help to reduce postoperative problems and improve the functional outcome. The main goal of this thesis is to develop and evaluate novel methods for personalized RSA, using state-of-the-art computer aided technologies to standardize and automate the design and planning phases.

Custom implants are a suitable solution when treating patients with extensive glenoid bone loss. However, clinical engineers are confronted with an enormous implant design space (number and type of screws, contact surface, etc.) and large anatomical and pathological variability. Currently, no objective tools exist to guide them when choosing the optimal design, i.e. with sufficient initial implant stability, thus making the design process tedious, time-consuming, and user-dependent. In this thesis, a Virtual Bench Test (VBT) simulation was developed using a finite element model to automatically evaluate the initial stability of custom shoulder implants. Through a validation experiment, it was shown that the virtual test bench output can be used by clinical engineers as a reference to support their decisions and adaptations during the implant design process.

When designing shoulder implants, knowledge about bone morphology and bone quality of the scapula throughout a certain population is fundamental. In particular, regions with the best bone stock (cortical bone) are taken into account to define the position and orientation of the screw holes, while aiming for an optimal fixation. As an alternative to manual measurements, whose generalization is limited by the analysis of small sub-sets of the potential patients, Statistical Shape Models (SSMs) have been commonly used to describe shape variability within a population. However, these SSMs typically do not contain information about cortical thickness. Therefore, a methodology to combine scapular bone shape and cortex morphology in an SSM was developed. First, a method to estimate cortical thickness, starting from a profile analysis of Hounsfield Unit (HU), was presented and evaluated. Then, using 32 manually segmented healthy scapulae, a statistical shape model including cortical information was created and assessed. The developed tool can be used to virtually implant a new design and test its congruency inside a generated virtual population, thus reducing the number of design iterations and cadaver labs.

Measurements of deltoid and rotator cuff muscle elongation during surgical planning can help clinicians to select a suitable implant design and position. However, such an assessment requires the indication of anatomical landmarks as a reference for the muscle attachment points, a process that is time-consuming and user-dependent, since often performed manually. Additionally, the medical images, which are normally used for shoulder arthroplasty, mostly contain only the proximal humerus, making it impossible to indicate those muscle attachment points which lie outside of the field of view of the scan. Therefore, a fully-automated method, based on SSM, for measuring deltoid and rotator cuff elongation was developed and evaluated. Its clinical applicability was demonstrated by assessing the performance of the automated muscle elongation estimation for a set of arthritic shoulder joints used for preoperative planning of RSA, thus confirming it a suitable tool for surgeons when evaluating and refining clinical decisions.

In this research, a major step was taken into the direction of a more personalized approach to Reverse Shoulder Arthroplasty, in which the surgical management, i.e. implant design and position, is adapted to the patient-specific characteristics and preoperative condition. By applying computer aided technologies in the clinical practice, design and planning process can be automated and standardized, thus reducing costs and lead times. Additionally, thanks to the novel methods presented in this thesis, we expect in the future a wider adoption of the personalized approach, with important benefits both for surgeons and patients.