ABSTRACT
PURPOSE: Establishing the symmetry of intraindividual orbital volumes is crucial for radiologic assessment, preoperative planning, and postoperative outcome evaluation. However, no reliable method exists to measure orbital volume because of problems in defining the bony boundaries of the orbit. Therefore, the purpose of this study was to propose a new approach to analyze human orbits and determine its application for quantifying bony symmetry in a cohort of patients. PATIENTS AND METHODS: Computed tomography scans of 93 patients were retrospectively collected from our institutional database. The intraindividual volume difference was quantified using a surface model derived from manual segmentations. The average shape of the orbit was calculated iteratively and nonrigidly registered to both orbits of all patients. After registration, the surface reconstructions of all orbits had an identical mesh topology and vertices at corresponding anatomic locations. The volume difference was calculated locally based on the relative position of the vertices at equivalent locations in the left orbit and right orbit. This approach was used to quantify the volume difference between the left and right orbits for all patients. Interobserver sensitivity was assessed in 5 randomly chosen patients and was measured independently by 3 specialists. RESULTS: An average difference of 600 ± 500 µL between the volumes of the left and right orbits was found, representing a difference of 2.1%. Although the difference in volume was small, the volumes were significantly different (P = .039). The largest asymmetries were found in the roof and floor area. CONCLUSIONS: The method proposed to measure the difference in volume between the left and right orbits is automated and does not rely on a closed orbital volume, which provides more objective volume measurements. With the help of modern computed tomography techniques and the coherent point drift method, it was possible to show that the intraindividual volume difference in the orbits is approximately 2%, not 7 to 8% as often cited in the literature.
