Risk assessment of dental-maxillofacial morphology affecting alveolar bone loss of the first molar in periodontitis based on a machine learning algorithm,a pilot study

Abstract

Abstract Background To establish a risk assessment of dental-maxillofacial morphology affecting alveolar bone loss in patients with periodontitis using machine learning algorithms.Methods Four machine learning algorithms were used to screen possible predictor variables such as age, sex, clinical probing depth (CPD), skeletal relationship between the maxilla and mandible (ANB angle), mandibular plane angle (FH-MP), upper and lower central incisor inclination (U1-L1). The algorithms were also used to establish a risk assessment model in patients with periodontitis. A receiver operating characteristic curve was used to evaluate the discrimination of the models. The model was visualized by a nomogram.Results The optimal variables screened were CPD and FH-MP using random forest algorithm; CPD, FH-MP and U1-L1 using lasso regression; and CPD, FH-MP, and age using both optimal subset regression and cross-validation. CPD, FH-MP, U1-L1, and age were selected as the optimal prediction subsets. The area under the receiver operating characteristic curve was 0.778.Conclusions Within the limitations of this study, FH-MP was an important predictor of the degree of alveolar bone loss in the first molar affected by periodontitis. The degree of alveolar bone loss in the first molar was more serious in high-angle periodontitis than in those with low- and average-angle periodontitis.