The adoption of statistical shape modeling into the realm of medical technologies has been explored with applications ranging from medical image analysis to registration techniques. Here, we have created 6 category-specific mandibular mesh templates, i.e. atlases, to enable subject-specific deformable registration methods. Our approach considers the characteristics of individuals including gender and dentition, compiling specific atlases based on these categories. Our objective is to improve deformable registration techniques through these category-dependent templates, which have not previously been investigated for the human mandible. We have evaluated non-rigid transformation techniques both with and without the use of the mandibular atlases and validated the results by comparison metrics between the surface of the resulting meshes and the patient model by Hausdorff distances and volumetric overlap scores. Our results showed no statistically significant difference between the average maximum Hausdorff 95 distance for cases using templates that matched the category of the test mandible when compared to templates that directly did not match F(2, 72) = 0.64, p = 0.53. The volumetric overlap scores offered more promising results where the matched group had a statistically significant greater mean than the unmatched group F(2, 33) = 12, p = 0.00012. Furthermore, the overlap percentage for matched cases (M = 74.08, SD = 8.91) was higher than unmatched cases (M = 59.42, SD = 6.05), t(22) = 4.72, p
CITATION STYLE
Borgard, H., Abdi, A. H., Prisman, E., & Fels, S. (2020). Creation of Categorical Mandible Atlas to Benefit Non-Rigid Registration. In Lecture Notes in Computational Vision and Biomechanics (Vol. 36, pp. 597–607). Springer. https://doi.org/10.1007/978-3-030-43195-2_50
Mendeley helps you to discover research relevant for your work.