Deep convolutional neural network-the evaluation of cervical vertebrae maturation

Area of Science:

• Dentistry
• Medical Imaging
• Artificial Intelligence

Background:

• Cervical vertebral maturation (CVM) assessment is crucial for orthodontic treatment planning.
• Manual CVM evaluation from lateral cephalometric radiographs can be subjective and time-consuming.

Purpose of the Study:

• To develop and evaluate a deep learning-based convolutional neural network (CNN) model for automated CVM stage determination.
• To assess the model's performance using precision, recall, and F1-score on lateral cephalometric radiographs.

Main Methods:

• A dataset of 588 digital lateral cephalometric radiographs from patients aged 8-22 years was utilized.
• A convolutional neural network (CNN) model was developed and trained using Python, Keras, and TensorFlow.
• The model's performance was evaluated based on CVM stage classification accuracy, precision, recall, and F1-score.

Main Results:

• The CNN model achieved a classification accuracy of 58.66% after 40 epochs of training.
• The model demonstrated similar performance on training (58%) and test (57%) datasets.
• Highest precision and F1-score were observed for CVM Stage 1, while highest recall was for CVM Stage 2.

Conclusions:

• The developed deep learning model shows moderate success in automatically classifying CVM stages from lateral cephalometric radiographs.
• The model's performance indicates potential for assisting clinicians in CVM assessment, though further refinement may be needed.