Computer-aided cephalometric landmark annotation for CBCT data.

Area of Science:

• Medical Imaging
• Computer-Aided Diagnosis
• Craniofacial Surgery

Background:

• Cone Beam Computerized Tomography (CBCT) is increasingly used in dental and maxillofacial practice.
• Manual cephalometric analysis, vital for diagnosis and treatment planning, suffers from landmark identification variability.
• 3D cephalometric analysis offers enhanced diagnostic capabilities but requires accurate landmark localization.

Purpose of the Study:

• To develop and validate a computer-aided approach for automatic 3D landmark annotation on CBCT data.
• To estimate the 3D positions of 21 selected cephalometric landmarks.
• To improve the accuracy and repeatability of cephalometric analysis compared to manual methods.

Main Methods:

• An adaptive cluster-based segmentation of bone tissues was performed.
• Intensity-based registration of an annotated reference volume to patient CBCT data was utilized.
• The algorithm was validated on 18 CBCT images, presenting results with confidence regions for manual refinement.

Main Results:

• Automatic segmentation demonstrated high accuracy with no significant difference from manual threshold values.
• The median localization error for landmarks was 1.99 mm (IQR: 1.22-2.89 mm).
• Segmentation proved robust, with accuracy comparable to existing methods.

Conclusions:

• The proposed computer-aided landmark annotation method is promising for 3D cephalometric analysis.
• The segmentation process is robust and achieves acceptable accuracy for most landmarks.
• This approach offers a viable alternative to manual landmark identification in CBCT analysis.