Quantification of Craniofacial Growth Pattern Based on Deep Learning.

Area of Science:

• Craniofacial development and growth analysis
• Artificial intelligence in medical imaging
• Quantitative radiology

Background:

• Childhood and adolescence are critical for craniofacial growth, impacting orthodontic and maxillofacial surgery decisions.
• Traditional cephalometric analysis is subjective and oversimplified; current AI methods often require manual annotations, limiting generalizability.
• Objective, comprehensive methods are needed to quantify complex craniofacial growth patterns.

Purpose of the Study:

• To develop an end-to-end deep learning framework for analyzing lateral cephalometric radiographs without manual annotations.
• To autonomously extract dynamic imaging features related to age intervals and sexual dimorphism in craniofacial development.
• To visualize and quantify spatiotemporal growth dynamics and sexual dimorphism using saliency maps and novel indices.

Main Methods:

• Utilized a deep learning framework on lateral cephalometric radiographs from 41,625 individuals (ages 4-18).
• Employed Gradient-weighted Class Activation Mapping (Grad-CAM) for feature visualization and generated population-averaged saliency maps.
• Introduced Age-related Saliency Index (ASI) and Sex-related Saliency Index (SSI) for quantitative evaluation of developmental and dimorphic characteristics.

Main Results:

• Age-related saliency maps highlighted internal bone details, visualizing regional importance during development, quantitatively prioritized by ASI.
• Sex-related Saliency Index (SSI) quantified sexual dimorphism evolution, showing early widespread differences concentrating in the mandible by adulthood.
• The framework successfully visualized and quantified dynamic craniofacial growth patterns and sex-specific characteristics.

Conclusions:

• An end-to-end deep learning framework was established for large-scale cephalometric radiograph analysis.
• Generated saliency maps and indices provide objective, quantitative references for assessing craniofacial developmental stages and sexual dimorphism.
• Findings offer novel insights into coordinated craniofacial bone growth and sex-specific radiological traits, aiding clinical decision-making.