Automatic Craniomaxillofacial Landmark Digitization via Segmentation-Guided Partially-Joint Regression Forest Model...

Area of Science:

• Medical Imaging
• Computer-Aided Diagnosis
• Biomedical Engineering

Background:

• Craniomaxillofacial (CMF) landmark digitization from cone-beam computed tomography (CBCT) images is crucial for diagnosis and treatment planning.
• Challenges include significant interpatient morphological variations and image artifacts inherent in CBCT data.
• Existing methods often struggle with accuracy and efficiency due to these complexities.

Purpose of the Study:

• To develop an automated system for accurate and efficient digitization of CMF landmarks from CBCT images.
• To address the challenges posed by patient-specific anatomical differences and CBCT image quality issues.
• To improve the reliability and consistency of landmark localization in CMF analysis.

Main Methods:

• Proposed a segmentation-guided partially-joint regression forest (S-PRF) model for automated CMF landmark digitization.
• Employed a regression voting strategy to aggregate contextual evidence for landmark localization, mitigating artifact impact.
• Utilized CBCT image segmentation to exclude irrelevant voxels and a partially-joint model for enhanced digitization reliability.
• Introduced a fast vector quantization method for efficient, low-dimensional, and artifact-invariant feature extraction.

Main Results:

• Achieved mean digitization errors below 2 mm for 15 CMF landmarks compared to ground truth.
• Demonstrated that the developed model effectively handles interpatient morphological variations and imaging artifacts.
• Experimental results on a CBCT dataset confirm clinically acceptable accuracy for landmark digitalization.

Conclusions:

• The proposed S-PRF model offers a robust solution for automated CMF landmark digitization from CBCT images.
• The method successfully overcomes limitations related to anatomical variability and image artifacts.
• This automated approach promises reduced labor costs and improved consistency in clinical CMF analysis.