Virtual reconstruction of orbital defects using Gaussian process morphable models.

Area of Science:

• Biomedical Engineering
• Medical Imaging
• Computer-Aided Surgery

Background:

• Conventional orbital defect reconstruction relies on manual mirroring, which is time-consuming and operator-dependent.
• Accurate reconstruction is crucial for designing patient-specific implants and restoring facial symmetry.

Purpose of the Study:

• To propose and evaluate an automated reconstruction method for orbital defects using a Gaussian process morphable model (GPMM).
• To compare the accuracy and efficiency of the GPMM method against the conventional mirroring technique.

Main Methods:

• A GPMM was created from 65 CT scans of healthy midfaces to model orbital shape variations.
• Parameter optimization involved quantitative metrics for generalization and specificity.
• Reconstruction error was assessed using artificial defects in 15 independent CT scans.

Main Results:

• The GPMM method achieved a lower median reconstruction error (0.35 ± 0.16 mm) compared to the mirroring method (0.52 ± 0.18 mm).
• Automated GPMM reconstruction demonstrated a median error of 0.39 ± 0.11 mm for large bilateral defects.
• The GPMM approach significantly reduced reconstruction time and improved accuracy.

Conclusions:

• GPMM-based reconstruction is more accurate and less time-consuming than the mirroring method for orbital defects.
• This automated technique shows significant potential as an alternative for patient-specific implant design.
• Further clinical validation is recommended to confirm its efficacy in patient populations.