Convolutional Neural Network-Based Deep Learning Methods for Skeletal Growth Prediction in Dental Patients

Abstract

This study aimed to predict the skeletal growth maturation using convolutional neural network-based deep learning methods using cervical vertebral maturation and the lower 2nd molar calcification level so that skeletal maturation can be detected from orthopantomography using multiclass classification. About 1200 cephalometric radiographs and 1200 OPGs were selected from patients seeking treatment in dental centers. The level of skeletal maturation was detected by CNN using the multiclass classification method, and each image was identified as a cervical vertebral maturation index (CVMI); meanwhile, the chronological age was estimated from the level of the 2nd molar calcification. The model’s final result demonstrates a high degree of accuracy with which each stage and gender can be predicted. Cervical vertebral maturation reported high accuracy in males (98%), while females showed high accuracy of 2nd molar calcification. CNN multiclass classification is an accurate method to detect the level of maturation, whether from cervical maturation or the calcification of the lower 2nd molar, and the calcification level of the lower 2nd molar is a reliable method to trust in the growth level, so the traditional OPG is enough for this purpose.

Keywords

• artificial intelligence
• multiclass classification
• growth maturation prediction
• lower 2nd molar