A retrospective study published in Frontiers in Medicine developed an XGBoost model to classify the cross-sectional severity of per-implant disease using only restoration features. The study evaluated 300 implants in 285 patients, achieving an internal validation accuracy of 78.9%, a macro-average AUC of 0.821, and a 95% confidence interval of 0.760 to 0.882, without using probing depth or the bleeding index. The findings indicate that the model may serve as an automated diagnostic aid for cross-sectional assessment, but it cannot be used to predict longitudinal outcomes.
Source:BIOENGINEER.ORG · Read the original
Leave a Reply