Overview

Skeletal Malocclusion Class III  (SCIII) is a complex craniofacial condition with highly variable morphology, making diagnosis and treatment planning challenging. This project explores how geometric morphometrics and machine learning can bring greater objectivity and precision to clinical decision-making.

By analyzing craniofacial shape data, we identify meaningful patient subgroups and develop predictive tools to support personalized treatment strategies.

Objectives

  • Identify clinically relevant subphenotypes of SCIII using data-driven clustering approaches
  • Develop machine learning models to classify patients into subphenotypes
  • Predict treatment strategy (surgical vs. orthodontic) based on craniofacial features

Data & Approach

  • Cohort: 655 adult SCIII patients
  • Data: Craniofacial landmarks, cephalometric measurements, and clinical records
  • Methods:
    • Geometric morphometrics for shape analysis
    • Unsupervised learning to uncover subgroups
    • Supervised models for classification and treatment prediction

Key Results

  • Identified six distinct craniofacial subphenotypes
  • Found strong links between subphenotypes and treatment decisions
  • Developed a predictive model that outperforms current approaches

Why It Matters

This work advances craniofacial diagnostics by:

  • Enabling more objective and reproducible patient classification
  • Supporting better patient stratification
  • Providing data-driven guidance for treatment planning

Outputs