Evaluation of automated photograph-cephalogram image integration using artificial intelligence models.

Area of Science:

• Medical Imaging
• Artificial Intelligence in Medicine
• Orthodontics

Background:

• Accurate superimposition of digital photographs and lateral cephalograms is crucial for diagnosis and treatment planning in orthodontics.
• Manual methods for image integration can be time-consuming and prone to variability.

Purpose of the Study:

• To develop and evaluate an automated method for combining digital photographs with lateral cephalograms using artificial intelligence (AI).

Main Methods:

• Developed two deep-learning AI models for soft tissue landmark detection on digital photographs and lateral cephalograms.
• Utilized image rotation, scaling, and shifting to align landmarks identified by AI models.
• Validated the automated method against manual image integration using paired t-tests on 100 subjects.

Main Results:

• The automated method showed no statistically significant difference compared to manual methods for most soft tissue landmarks.
• Minor statistically significant differences were observed for the upper lip and soft tissue B point.

Conclusions:

• Automated photograph-cephalogram image integration using AI models is a reliable alternative to manual superimposition procedures.
• This AI-driven approach offers potential for increased efficiency and consistency in clinical practice.