OCLU-NET for occlusal classification of 3D dental models

• Published in: Machine vision and applications Vol. 31; no. 6
• Main Authors: Juneja, Mamta, Singla, Ridhima, Saini, Sumindar Kaur, Kaur, Ravinder, Bajaj, Divya, Jindal, Prashant
• Format: Journal Article
• Language: English
• Published: Berlin/Heidelberg Springer Berlin Heidelberg 01.09.2020; Springer Nature B.V
• Subjects: Artificial neural networks; Classification; Communications Engineering; Computer Science; Deep learning; Dental materials; Dentistry; Histograms; Image classification; Image Processing and Computer Vision; Lithography; Machine learning; Networks; Occlusion; Pattern Recognition; Special Issue on Deep Learning Methods for Biomedical Information Analysis; Special Issue Paper; Three dimensional models; Vision systems; Occlusion; Deep learning; Random forest regressor; K-nearest neighbours; Support vector machine; Convolutional neural network
• ISSN: 0932-8092; 1432-1769
• DOI: 10.1007/s00138-020-01102-4

With the emergence in modern dentistry, the study of dental occlusion has been a subject of major interest. The aim of the present study is to investigate the capabilities of deep learning for the classification of dental occlusion using 3D images that has an exciting impact in several fields of dental anatomy. In present work, the 3D stereolithography (STL) files depicting the dental structures are converted to 2D histograms, using Absolute Angle Shape Distribution (AAD) technique, which are used as an input to deep or machine learning models for classification of dental structures based on the similarity of their shape features. To the best of the authors’ knowledge, no solution has been proposed for classification of dental occlusion using deep learning. Thus, an attempt has been made to propose a classification technique for dental occlusion. Based on the experimental analysis, it has been revealed that the deep learning-based convolutional neural network along with AAD performs better as compared to other existing machine learning techniques, with maximum accuracy of 78.95% for occlusion classification. However, the presented study is preliminary, but the experimental outcomes have demonstrated that deep learning is helpful in classifying dental occlusion and it has great application potential in the computer-assisted orthodontic treatment diagnosis.