Total elbow arthroplasty has become a clinically accepted treatment option for patients with disorders of the elbow. Implant malpositioning, as a result of an incorrect reproduction of the elbow's flexion-extension axis, is an important cause of implantrelated complications. Current surgical approaches introduce axis estimation errors that can vary significantly, and can be magnified in the presence of compromised bone. This work aimed to develop and evaluate image-based landmark selection and registration techniques to improve the accuracy of axis identification and thus, implant placement;​ with a focus on the distal humerus. A preliminary registration study was conducted, evaluating three registration techniques. Registration accuracy was highest using a surface-based registration combined with a non-contact surface acquisition device. The effect of anatomic landmark selection on registration accuracy was then investigated. An accurate registration was achievable using landmarks that are readily available to the surgeon and unlikely compromised due to humeral bone loss. An anthropometric study of the distal humerus investigated differences between paired specimens. Side-to-side measurements were similar enough that the contralateral anatomy may be referenced in the absence of anatomic landmarks due to bone loss. These techniques were applied to an image-based navigation system, and the accuracy of navigated implant alignment was compared against a non-navigated approach. The effect of implant design was also evaluated by comparing the alignment of a standard commercial humeral implant component against an implant with a reduced humeral stem. Implant alignment was significantly improved for the navigated approach. Implant alignment without navigation relied heavily on bone quality, whereas navigated alignment was consistent, regardless of bone quality. Further, a fixed articulation axis angle that is standard with commercial implants significantly affected the ability to correctly position the implant under navigation.

In summary, these studies introduced image-based landmark selection and registration techniques, quantified morphological differences between paired distal humeri, identified the sources of error attributed to implant positioning in a conventional surgical setting and identified the primary design constraint of modern commercial implants. The results of this dissertation will contribute to improved surgical procedures and implant design, improving long-term clinical outcomes.