An Adaptive Region Growing Method to Segment Inferior Alveolar Nerve Canal from 3D Medical Images for Dental Implant Surgery

Abstract

Digital shape reconstruction is the process of creating digital models from physical parts represented by 3D point clouds. The ideal process is expected to provide a boundary representation that is likely to be identical or similar to the original design intent of the object, and requires minimal user assistance. This paper discusses alternative state-of-the-art approaches, where emphasis is put on automatic methods (i) to create complete and consistent topological structures over polygonal meshes; and (ii) extract accurate and properly aligned surface features that yield complete, trimmed CAD models with fillets and corner patches. Problems and recommended solutions will be presented through case studies using industrial parts.

Automatic Extraction of Mandibular Nerve and Bone from Cone-Beam CT Data

“…It yields an accuracy of 1.6 mm for the cortical bone, and 3.4 mm for the dental nerve on 215 single 2D slices in the CT data. More recently, Yau et al [5] proposed a semi-automatic method to segment the nerve canal from conventional CT data. It requires the user to manually specify a seed point for a subsequent automatic adaptive region-growing approach in consecutive slices of the CT data.…”

Accuracy of software designed for automated localization of the inferior alveolar nerve canal on cone beam CT images

“…In the first step, panoramic data must be derived as described elsewhere 4,12,14 (Figure 1). By applying the active contour technique on panoramic projections, the Automated localization of the IAN canal on CBCT images Bahrampour et al mandibular bone is separated, and redundant points are removed ( Figure 2).…”

“…Table 1 summarizes the existing methods. 1,2,4,[7][8][9][10][11][12][13][14][15][16] To the best of our knowledge, only three studies focused on CBCT data and presented a fully automated method as well. 2,7,15 In one of these studies, Kainmueller et al 7 (at the University Hospital, Cologne, Germany, in 2009) segmented the mandible using an active shape model (ASM), which is constructed from 106 clinical data sets.…”

Segmentation of the Mandibular Canal in Cone-Beam CT Data

“…2008 "An adaptive region growing method to segment inferior alveolar nerve canal from 3D medical images for dental implant surgery", by Yau et al [32], first reslices the image volume, to a kind of panoramic image volume. Then an initial point is selected by the user inside the canal, followed by local region growing and automatic selection of a new initial point for region growing.…”

“…The literature describes several methods to segment, automatically or user assisted, the mandibular canal in CT. Methods based on fast marching shortest path tracing [28], region growing [32] and segmentation methods based on image gradients [29]. We have tested these methods on our CBCT data, however, all fail because of low contrast between canal and surrounding tissue, high noise and missing edges (figure 4.1).…”