Determining 3D orientation of irregularly shaped objects from biplane 2D images
Nicholson, Kristen Faith
University of Delaware
In this study, static bi-plane radiographs were collected for 30 unique scapula orientations, using 10 different scapulae. Static images were used because of ease of access to the imaging system, but the matching approach can be extended to fluoroscopic images collected during shoulder motion. The matching technique was tested with both a scapula specific 3D model and a scaled general 3D model. The ability to use a general model would decrease the dependency on imaging hardware. Accordingly, the ability to use a general model would facilitate implementing the approach on a larger scale. When using the scapula specific model, the errors ranged from 2.46° to 10.21°. The scaled general model yielded errors between 2.43° and 13.1°. Statistically, only the comparison of the resolved Y positions for the matches obtained with the scapula specific models and the motion capture data were significantly different. The developed matching technique was shown to be accurate with both the scapula specific model and the general scaled model. The orientations represented the entire range of motion of a scapula, so this same technique can be considered plausible for matching scaled general models to bi-plane fluoroscopic images of motion data. Since the analysis of 2D fluoroscopic images involves radiation exposure and is time consuming, this technique is more likely to serve as the gold standard for techniques that use more clinically available measures, rather than serving as a clinical tool itself.