Tania Camila Niño-Sandoval1, Sonia V Guevara Pérez2, Fabio A González3, Robinson Andrés Jaque4, Clementina Infante-Contreras5. 1. Universidad Nacional de Colombia - Bogotá, Faculty of Dentistry, Oral Health Department, Master of Dentistry, Craniofacial Growth and Development Research Group, Genetics Institute, Cll 53-Cra. 37 Ed. 426 Of. 213, Bogotá, Colombia. Electronic address: kotc2578@gmail.com. 2. Universidad Nacional de Colombia - Sede Bogotá, Faculty of Dentistry, Oral Health Department-Orthodontics, Craniofacial Growth and Development Research Group, 11001 Bogotá, Colombia. Electronic address: svguevarap@unal.edu.co. 3. Universidad Nacional de Colombia - Bogotá, Faculty of Engineering, Computing Systems and Industrial Engineering Department, MindLab Research Group, Carrera 30 N° 45-03, Bogotá, Colombia. Electronic address: fagonzalezo@unal.edu.co. 4. Universidad Nacional de Colombia - Bogotá, Faculty of Engineering, Computing Systems and Industrial Engineering Department, MindLab Research Group, Carrera 30 N° 45-03, Bogotá, Colombia. Electronic address: rajaquep@unal.edu.co. 5. Universidad Nacional de Colombia - Bogotá, Faculty of Dentistry, Oral Health Department, Master of Dentistry, Craniofacial Growth and Development Research Group, Genetics Institute, Cll 53-Cra. 37 Ed. 426 Of. 213, Bogotá, Colombia. Electronic address: ccontrerasi@unal.edu.co.
Abstract
BACKGROUND: The prediction of the mandibular bone morphology in facial reconstruction for forensic purposes is usually performed considering a straight profile corresponding to skeletal class I, with application of linear and parametric analysis which limit the search for relationships between mandibular and craniomaxillary variables. OBJECTIVE: To predict the mandibular morphology through craniomaxillary variables on lateral radiographs in patients with skeletal class I, II and III, using automated learning techniques, such as Artificial Neural Networks and Support Vector Regression. MATERIALS AND METHODS: 229 standardized lateral radiographs from Colombian patients of both sexes aged 18-25 years were collected. Coordinates of craniofacial landmarks were used to create mandibular and craniomaxillary variables. Mandibular measurements were selected to be predicted from 5 sets of craniomaxillary variables or input characteristics by using automated learning techniques, and they were evaluated through a correlation coefficient by a ridge regression between the real value and the predicted value. RESULTS: Coefficients from 0.84 until 0.99 were obtained with Artificial Neural Networks in the 17 mandibular measures, and two coefficients above 0.7 were obtained with the Support Vector Regression. CONCLUSION: The craniomaxillary variables used, showed a high predictability ability of the selected mandibular variables, this may be the key to facial reconstruction from specific craniomaxillary measures in the three skeletal classifications.
BACKGROUND: The prediction of the mandibular bone morphology in facial reconstruction for forensic purposes is usually performed considering a straight profile corresponding to skeletal class I, with application of linear and parametric analysis which limit the search for relationships between mandibular and craniomaxillary variables. OBJECTIVE: To predict the mandibular morphology through craniomaxillary variables on lateral radiographs in patients with skeletal class I, II and III, using automated learning techniques, such as Artificial Neural Networks and Support Vector Regression. MATERIALS AND METHODS: 229 standardized lateral radiographs from Colombian patients of both sexes aged 18-25 years were collected. Coordinates of craniofacial landmarks were used to create mandibular and craniomaxillary variables. Mandibular measurements were selected to be predicted from 5 sets of craniomaxillary variables or input characteristics by using automated learning techniques, and they were evaluated through a correlation coefficient by a ridge regression between the real value and the predicted value. RESULTS: Coefficients from 0.84 until 0.99 were obtained with Artificial Neural Networks in the 17 mandibular measures, and two coefficients above 0.7 were obtained with the Support Vector Regression. CONCLUSION: The craniomaxillary variables used, showed a high predictability ability of the selected mandibular variables, this may be the key to facial reconstruction from specific craniomaxillary measures in the three skeletal classifications.
Keywords:
Artificial Neural Networks; Forensic anthropology population data; Mandibular prediction; Skeletal class I, II, III malocclusion; Support Vector Regression
Authors: Shankargouda Patil; Sarah Albogami; Jagadish Hosmani; Sheetal Mujoo; Mona Awad Kamil; Manawar Ahmad Mansour; Hina Naim Abdul; Shilpa Bhandi; Shiek S S J Ahmed Journal: Diagnostics (Basel) Date: 2022-04-19
Authors: Yashodhan M Bichu; Ismaeel Hansa; Aditi Y Bichu; Pratik Premjani; Carlos Flores-Mir; Nikhilesh R Vaid Journal: Prog Orthod Date: 2021-07-05 Impact factor: 2.750