Accuracy of machine learning-assisted prediction of the future need for orthognathic surgery in patients with cleft lip and palate.

Objective: To investigate the accuracy of machine learning (ML)-assisted prediction of the need for orthognathic surgery (OGS) in patients with cleft lip and palate (CLP). Methods: This study included 245 patients with CLP whose lateral cephalograms were available at pre-adolescence (T0; mean age, 8.45 years) and young adulthood (T1; mean age: 18.37 years). At T1, the patients were classified into the surgery group based on two criteria: (1) satisfying at least three of the following four conditions: ANB < -3°, Wits appraisal < -5 mm, APDI > 90°, and AB-MP < 60° and (2) undergoing presurgical orthodontic treatment or having undergone OGS. A total of 25.3% (n = 62) of patients were assigned to the surgery group, while 74.7% (n = 183) were assigned to the non-surgery group. Further, 80% and 20% of each group were used as training/validation and test sets, respectively. After 37 cephalometric variables and two cleft-related variables were measured, support vector machine (SVM) and feature importance analysis (FIA) with Shapley additive explanation were used to determine the prediction accuracy and predictors at T0. Results: SVM demonstrated area under curve 0.84, accuracy 83.7%, sensitivity 83.3%, and specificity 83.8%. FIA revealed 10 predictors: A to N-perpendicular, L1 to A-Pog, Pog to N-perpendicular, L1 to Lower-occlusal plane, Cleft type, U1 to Upper-occlusal plane, IMPA, gonial angle, anteroposterior facial height ratio, and ANB with accumulated importance of 64.51%. Conclusions: The ML algorithm used in this study may support clinical decision-making in identifying candidates for future OGS at 8 years of age.