OBJECTIVE: This study aimed to design an artificial neural network for the prediction of cephalometric variables via a lateral photo- graph in skeletal Class I, II, and III patterns. METHODS: A total of 94 patients were recruited for this prospective study, with an age range of 15-20 years (41 boys and 53 girls) seek- ing orthodontic treatment. According to cephalometric analysis, using AutoCAD 21.0, they were allocated into three groups. Thirty with skeletal Class I (14 boys and 16 girls), 34 with skeletal Class II (14 boys and 20 girls), and 30 with skeletal Class III malocclusion (13 boys and 17 girls) according to SNA, SNB, and ANB angles measured from cephalometric radiographs. The study includes (1) finding the correlation of the skeletal measurements between lateral profile photographs and cephalometric radiographs for the recruited patients and (2) designing a specific artificial neural networks for the assessment of skeletal factors via lateral photographs, these artificial neural networks are trained and tested with the total of 94 standard lateral cephalograms. RESULTS: This novel Network provided models of regression that can forecast the cephalometric variables through analogous photo- graphic measurements with excellent predictive power R = 0.99 and limited estimation error for each malocclusion (Class I, II, and III). CONCLUSION: This study suggests that artificial intelligence would be useful as an accurate method in orthodontics for the prediction of cephalometric variables and its performance was achieved by several factors such as proper selection of the input data, preferable generalization, and organization.
OBJECTIVE: This study aimed to design an artificial neural network for the prediction of cephalometric variables via a lateral photo- graph in skeletal Class I, II, and III patterns. METHODS: A total of 94 patients were recruited for this prospective study, with an age range of 15-20 years (41 boys and 53 girls) seek- ing orthodontic treatment. According to cephalometric analysis, using AutoCAD 21.0, they were allocated into three groups. Thirty with skeletal Class I (14 boys and 16 girls), 34 with skeletal Class II (14 boys and 20 girls), and 30 with skeletal Class III malocclusion (13 boys and 17 girls) according to SNA, SNB, and ANB angles measured from cephalometric radiographs. The study includes (1) finding the correlation of the skeletal measurements between lateral profile photographs and cephalometric radiographs for the recruited patients and (2) designing a specific artificial neural networks for the assessment of skeletal factors via lateral photographs, these artificial neural networks are trained and tested with the total of 94 standard lateral cephalograms. RESULTS: This novel Network provided models of regression that can forecast the cephalometric variables through analogous photo- graphic measurements with excellent predictive power R = 0.99 and limited estimation error for each malocclusion (Class I, II, and III). CONCLUSION: This study suggests that artificial intelligence would be useful as an accurate method in orthodontics for the prediction of cephalometric variables and its performance was achieved by several factors such as proper selection of the input data, preferable generalization, and organization.