The aim of this study was to define axes from clearly identifiable landmarks on the proximal aspect of the humerus and to compare these for reasonable best alternatives to the use of the humeral canal and elbow epicondylar axes to define a humeral coordinate frame (HCF). The elbow epicondylar axis (EC) and six different humeral canal axes (HC) based on varying lengths of humerus were quantified from 21 computed tomography (CT) scans of humeri. Six additional axes were defined using the proximal humerus only. These included a line from the center of a sphere fit on the humeral head to the 3D surface area centroid of the greater tubercle region, (GT). The inclinations of these axes relative to EC were calculated. GT was found to be the most closely aligned to EC (13.4° ± 6.8°). The inclinations of the other axes ranged from 36.3° to 86.8°. The HC axis orientation was found to be insensitive to humeral shaft lengths (variability, within average: 0.6°). This was chosen as one of two axes for the HCF. It was also the most inter-subject related axis to EC with inclination standard deviation of ±1.8°. EC was therefore predicted from this such that if the superior axis [1 0 0] of an image scan is maintained and the humerus rotated to make its quantified HC align superiorly in the direction [0.98 0.01 0.01], then its EC axis lies laterally in the direction [0 0 1]. This study demonstrates that it is possible with confidence to apply an orthogonal coordinate frame to the humerus based on proximal imaging data only