Lizhe Xie1,2, Wen Tang3,2, Iman Izadikhah3,2, Xiaoyu Chen3,2, Zhenqi Zhao4, Yang Zhao5, Hu Li3,2, Bin Yan6,7,8. 1. Jiangsu Province Key Laboratory of Oral Diseases, Nanjing Medical University, Nanjing, China. 2. Jiangsu Province Engineering Research Center of Stomatological Translational Medicine, Nanjing Medical University, 136 Hanzhong Street, Gulou District, Nanjing, 210029, China. 3. Department of Orthodontics, The Affiliated Stomatological Hospital of Nanjing Medical University, 136 Hanzhong Street, Gulou District, Nanjing, 210029, China. 4. Department of Stomatology, The First People's Hospital of Nantong, Nantong, China. 5. Department of Biostatistics, School of Public Health, Nanjing Medical University, Nanjing, China. 6. Department of Orthodontics, The Affiliated Stomatological Hospital of Nanjing Medical University, 136 Hanzhong Street, Gulou District, Nanjing, 210029, China. byan@njmu.edu.cn. 7. Jiangsu Province Key Laboratory of Oral Diseases, Nanjing Medical University, Nanjing, China. byan@njmu.edu.cn. 8. Jiangsu Province Engineering Research Center of Stomatological Translational Medicine, Nanjing Medical University, 136 Hanzhong Street, Gulou District, Nanjing, 210029, China. byan@njmu.edu.cn.
Abstract
OBJECTIVE: To develop new logistic regression estimative models of the cervical vertebral maturation index (CVMI) based on cone-beam CT (CBCT)-derived parameters for intelligent evaluating skeletal maturation. METHODS: From 231 CBCT volumes (age range 7-17, mean age 11.09 years), 154 were randomly selected to produce 2D sagittal projections of the second to fourth cervical vertebrae (C2-C4). From 19 quantitative parameters, significant predictors were deduced to formulate logistic models. Using the CVMI and significant predictors of 77 other subjects, performance of the models was externally examined by direct comparison and the area under the receiver operating characteristic curve (AUC). Models were modified if required, to improve their accuracy. RESULTS: Chronological age, C3 height ([Formula: see text], and ratio of posterior height to lower width of C4 [Formula: see text] were entered as significant predictors. Accuracy of the models was acceptable (total AUC = 0.91) except for 4th and 5th stage (AUC of 0.82 and 0.83, respectively), which were mis-predicted inversely. Adjusted models were generated by bivariate logistic regression analysis and adding significant parameters ([Formula: see text] and [Formula: see text], with odds ratios of 3.308 and 3.38, respectively) from 58 subjects in 4th and 5th stages of CVMI in the model establishment group. The total AUC increased to 0.94, along with an increase in the accuracy of the latter optimized models to 77.9 and 87%, respectively. CONCLUSION: The new intelligent models reliably estimated skeletal maturation and can be utilized in the clinical field or machine learning-based skeletal maturation assessment.
OBJECTIVE: To develop new logistic regression estimative models of the cervical vertebral maturation index (CVMI) based on cone-beam CT (CBCT)-derived parameters for intelligent evaluating skeletal maturation. METHODS: From 231 CBCT volumes (age range 7-17, mean age 11.09 years), 154 were randomly selected to produce 2D sagittal projections of the second to fourth cervical vertebrae (C2-C4). From 19 quantitative parameters, significant predictors were deduced to formulate logistic models. Using the CVMI and significant predictors of 77 other subjects, performance of the models was externally examined by direct comparison and the area under the receiver operating characteristic curve (AUC). Models were modified if required, to improve their accuracy. RESULTS: Chronological age, C3 height ([Formula: see text], and ratio of posterior height to lower width of C4 [Formula: see text] were entered as significant predictors. Accuracy of the models was acceptable (total AUC = 0.91) except for 4th and 5th stage (AUC of 0.82 and 0.83, respectively), which were mis-predicted inversely. Adjusted models were generated by bivariate logistic regression analysis and adding significant parameters ([Formula: see text] and [Formula: see text], with odds ratios of 3.308 and 3.38, respectively) from 58 subjects in 4th and 5th stages of CVMI in the model establishment group. The total AUC increased to 0.94, along with an increase in the accuracy of the latter optimized models to 77.9 and 87%, respectively. CONCLUSION: The new intelligent models reliably estimated skeletal maturation and can be utilized in the clinical field or machine learning-based skeletal maturation assessment.