A semi-automatic approach for longitudinal 3D upper airway analysis using voxel-based registration.

Area of Science:

• Medical Imaging
• Radiology
• Anatomy

Background:

• Accurate three-dimensional (3D) analysis of the upper airway (UA) is crucial for diagnosing and managing conditions like obstructive sleep apnea and planning surgical interventions.
• Traditional methods for UA analysis often rely on manual measurements, which can be time-consuming and prone to inter-observer variability.

Purpose of the Study:

• To introduce and validate a novel, semi-automatic approach for 3D upper airway analysis utilizing voxel-based registration (VBR).
• To assess the reliability and accuracy of this VBR-based method for quantifying UA volumes and cross-sectional areas (CSA) in post-orthognathic surgery patients.

Main Methods:

• Post-operative cone beam computed tomography (CBCT) scans were registered to pre-operative scans using VBR, referencing the anterior cranial base.
• Anatomical landmarks facilitated automatic segmentation of the UA for volume and CSA calculations.
• Intra- and inter-observer reliability was evaluated using intraclass correlations and Bland-Altman analysis, with comparisons to a manual landmark-based method.

Main Results:

• The VBR method demonstrated excellent intra- and inter-observer reliability (Intraclass Correlation Coefficient [ICC] ≥ 0.995).
• Bland-Altman plots indicated good observer agreement, minimal bias, and no systematic errors.
• The relative Dahlberg error for volume ranged from 0.51-4.30% and for CSA from 0.24-2.90%, showing improvement over manual methods. Method error introduced by VBR was 0.05-1.44%.

Conclusions:

• The proposed semi-automatic VBR method for 3D upper airway analysis is highly reliable and achieves excellent observer agreement.
• This approach is feasible for large-scale longitudinal clinical studies due to its semi-automatic nature, offering efficiency and consistency.