Clinical-imaging Model for Predicting Prognosis in Contemporary Endodontic Microsurgery: A Retrospective Machine...

Area of Science:

• Dentistry
• Biomedical Engineering

Background:

• Predictive models for endodontic microsurgery (EMS) outcomes are limited.
• This study addresses the need for improved prognostic tools in EMS.

Purpose of the Study:

• To evaluate machine learning (ML) algorithms for predicting EMS prognosis.
• To identify key patient, tooth, and procedure-related predictors of EMS outcomes.

Main Methods:

• Retrospective analysis of 213 teeth using ML algorithms.
• Data preprocessing included SMOTE for class imbalance and SelectKbest for feature selection.
• Classifiers like random forest were trained and validated; SHapley Additive exPlanations (SHAP) were used for interpretation.

Main Results:

• The random forest classifier achieved high predictive performance (accuracy: 91%, AUC: 0.97).
• Identified predictors of poor prognosis include lack of guided tissue regeneration, poor root-end filling quality, and lesion size.

Conclusions:

• The random forest model shows strong predictive potential for EMS outcomes.
• SHAP-derived predictors offer clinical insights but require external validation for definitive conclusions.