Development and validation of an artificial intelligence algorithm for cervical vertebral maturation staging using...

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Orthodontics

Background:

• Cervical vertebral maturation (CVM) assessment via lateral cephalograms is crucial for evaluating skeletal maturity without extra radiation.
• Traditional manual CVM analysis is time-consuming and prone to inter-observer variability.
• AI in medical imaging presents an opportunity to enhance diagnostic efficiency and accuracy.

Purpose of the Study:

• To develop and validate a deep learning algorithm for automated CVM staging.
• To improve the diagnostic efficiency and accuracy of skeletal maturity assessment.
• To leverage AI for analyzing lateral cephalometric radiographs.

Main Methods:

• Utilized 525 lateral cephalograms from individuals aged 7-17 years.
• Employed an AI platform (PLAINSIGHT) for landmark identification and vertebral measurements.
• Trained a VGG19 convolutional neural network model on 1300 augmented images and validated on 105 independent cephalograms.

Main Results:

• The VGG19 model achieved 86% overall accuracy in CVM staging.
• Optimal performance was noted between 80-100 training epochs.
• High classification accuracy was observed for CVS stages 4, 5, and 6, with an overall F1 score of 0.85.

Conclusions:

• The VGG19 deep learning model demonstrates significant potential for automating CVM assessment.
• The AI model offers high accuracy and reproducibility for evaluating skeletal development.
• This AI tool can serve as a valuable clinical decision-support system for growing individuals.