Generalized Automated Plate Computation for Cleft Lip and Palate

Ruben Schenk – January 2025

Project Overview

This project aimed to enhance a fully automated pipeline for presurgical orthopedic plate design for newborns with cleft lip and palate (CLP). Conducted in collaboration with the University Hospital Basel and the University of Basel, the focus was on improving plate design for bilateral cleft lip and palate (BCLP) cases, extending buccal areas, and stabilizing the premaxilla.

Background

Cleft lip and palate are common congenital craniofacial anomalies, affecting approximately 0.33 to 0.45 per 1000 live births globally. Presurgical orthopedic (PSO) plates are essential for aligning alveolar ridges and reducing cleft size, thus facilitating subsequent surgical interventions.

Existing automated pipelines provide robust solutions for unilateral CLP (UCLP) cases but are limited in handling more complex anatomical variations in BCLP and other rare cleft types. The focus of this semester thesis was to address these limitations by enhancing three key components of the pipeline:

  1. Improved Registration using Non-Rigid Iterative Closest Point (NICP)
  2. Buccal Area Extension via Template Deformation
  3. Premaxilla Inclusion in BCLP Cases

Methodology

1. Improved Registration with NICP

The original implementation of NICP led to plate shrinkage in the buccal area, particularly in cases with incomplete scan data. The modified algorithm incorporated an angular constraint:

\[ \cos(\theta) = \frac{c_i \cdot n_i}{||c_i|| ||n_i||} \]

where:

  • \( c_i \) is the connecting vector from the template vertex to the target point,
  • \( n_i \) is the surface normal at the target point.

The angular constraint filters correspondences that deviate significantly in alignment, preventing undesired deformations and maintaining anatomical accuracy.

2. Buccal Area Extension

To enhance the fit of the plate in BCLP cases, the template was dynamically extended in the buccal region using the As-Rigid-As-Possible (ARAP) deformation method. The extension process involved defining two key vertex categories:

  • Handle Vertices: Boundary vertices along the buccal region, which are displaced outward by a user-defined amount.
  • Stiff Vertices: Vertices fixed to preserve anatomical structure and prevent excessive deformation.

3. Premaxilla Inclusion

The premaxilla (PM) was integrated into the plate design for BCLP cases. This involved modifying the plate index set and introducing a pre-smoothing step to reduce artefacts.

Results

  • Improved NICP: The adapted NICP algorithm successfully mitigated buccal area shrinkage, achieving 0.5mm to 1mm of additional coverage.
  • Buccal Area Extension: The 1mm extension improved plate fit without introducing artefacts, whereas the 2mm extension occasionally misaligned with scan artefacts.
  • Premaxilla Inclusion: Integrating the PM stabilized the plate in BCLP cases, but further refinements are needed for cases with significant deformities.

Conclusion and Outlook

The enhanced pipeline demonstrated significant improvements in registration robustness, buccal area adaptability, and premaxilla inclusion. Future work includes refining the PM inclusion approach using localized smoothing.

For the complete report, please access the full semester thesis here.