Histogram - Based Algorithm for Semiautomated Three-Dimensional Craniofacial Modeling
MetadataShow full item record
Volume averaging artifacts in medical imaging result from voxel occupancy by more than one tissue type and, with anisotropic voxels, may be decreased by changing the imaging plane orientation relative to the target tissue and/or by decreasing slice thickness.1 In craniofacial CT imaging, volume averaging artifact becomes significant in areas ofthin bone such as the orbital walls and auditory ossicles. These regions are customarily imaged using multiple scan planes and the thinnest slices possible to reduce such artifacts. In three-dimensional craniofacial imaging, these same parameters may be controlled to reduce partial volume averaging, but areas of bone "drop-out" (also called pseudoforamina) are commonly present secondary to a paradoxical inability to lower thresholds without including unwanted background tissues. At present, the optimal solution to this problem is achieved by manually (and often painstakingly) drawing a region of interest around tissues presumed to contain volume averaged target density voxels and lowering thresholds to include these voxels in the 3D reconstruction, one CT slice at a time. Recently, anatomic modeling technologies have demonstrated the feasibility of assembling particulate hydroxyapatite (SYnthetic bone) into detailed craniofacial models of high anatomic accuracy, theoretically suitable for in vivo implantation (work in progress with the Department of Mechanical Engineering, University of Texas at Austin.) These modeling systems, such as stereolithography and selective laser sintering, operate as do 3D imaging workstations, using thresholds to 198 include/exclude pixels from CT data sets in the modeling process. However, the user interactive capabilities ofsuch technologies may be limited such that manual tracing ofregions of volume averaged thin bone is not possible. Drop-out artifacts in models so generated would be potentially larger than on corresponding 3D images where user input could reduce these artifacts. To circumvent this inability to manually correct volume averaging artifacts on anatomic modeling systems, and to relieve the intensive oPerator input required to otherwise achieve this goal on 3D imaging software, a histogram-based algorithm for semiautomated threedimensional craniofacial modeling was develoPed.