Cephalometric Variables Prediction from Lateral Photographs Between Different Skeletal Patterns Using Regression...

Area of Science:

• Orthodontics
• Artificial Intelligence
• Medical Imaging

Background:

• Cephalometric analysis is crucial for diagnosing and planning orthodontic treatment.
• Predicting cephalometric variables from lateral photographs can streamline diagnosis.

Purpose of the Study:

• To design an artificial neural network (ANN) for predicting cephalometric variables using lateral photographs.
• To assess the accuracy of ANN in classifying skeletal patterns (Class I, II, III).

Main Methods:

• 94 orthodontic patients (15-20 years) were classified into skeletal Class I, II, and III based on cephalometric analysis.
• Lateral cephalograms and profile photographs were used to train and test the ANN.
• The ANN was designed to correlate skeletal measurements from photographs with cephalometric radiographs.

Main Results:

• The developed ANN demonstrated excellent predictive power (R=0.99) for cephalometric variables.
• The model showed limited estimation error across all skeletal malocclusion classes.
• ANN accurately forecasted cephalometric variables from analogous photographic measurements.

Conclusions:

• Artificial intelligence offers a highly accurate method for cephalometric variable prediction in orthodontics.
• Proper input data selection, generalization, and organization are key to ANN performance.
• This AI approach can aid in orthodontic diagnosis and treatment planning.