A Novel Registration-Based Semiautomatic Mandible Segmentation Pipeline Using Computed Tomography Images to Study...

Area of Science:

• Medical imaging analysis
• 3D reconstruction
• Biomedical engineering

Background:

• Accurate segmentation of 3D mandibles from computed tomography (CT) data is crucial for growth studies.
• Manual segmentation is time-consuming and limits the scale of research.
• A need exists for efficient and automated methods for mandible segmentation.

Purpose of the Study:

• To develop and present a registration-based semiautomatic mandible segmentation (SAMS) pipeline.
• To enable efficient processing of large volumes of CT studies for 3D mandible segmentation.
• To reduce the manual effort required for mandible segmentation.

Main Methods:

• The SAMS pipeline involves manual preprocessing, automatic segmentation using nonlinear diffeomorphic registration, and manual postprocessing.
• Automatic segmentation registers CT scans to 54 age-specific templates (birth to 19 years).
• A composite 3D mandible is generated by combining 54 individual segmentations.

Main Results:

• The SAMS pipeline achieved 97% similarity agreement with manual segmentation for 20 test cases.
• Volumes of SAMS-processed mandibles were comparable to manually segmented ones.
• Manual postprocessing refined specific regions of the segmented 3D mandibles.

Conclusions:

• The SAMS pipeline is a robust method for 3D mandible segmentation from CT scans.
• It reduces manual segmentation time by 75%.
• The pipeline enhances the feasibility of conducting large-scale mandibular growth studies.