AI Prognostic Models in Endodontics: Scoping Review and Clinical Gap Analysis for Retreatment vs Microsurgery Decision Support.

INTRODUCTION

Artificial intelligence (AI) prognostic models in endodontics may estimate treatment outcomes, but single-pathway prediction does not provide comparative decision support for failed cases in which nonsurgical retreatment (NS-ReTx) and endodontic microsurgery (EMS) may both be considered. This scoping review mapped AI prognostic models in endodontics and evaluated their alignment with prespecified clinical decision domains relevant to NS-ReTx-versus-EMS selection.

METHODS

A dual-track scoping review was conducted according to PRISMA-ScR. PubMed, Embase, Web of Science, Cochrane CENTRAL, IEEE Xplore, ACM Digital Library, and arXiv were searched from inception through February 23, 2026. Track 1 identified AI models developed to estimate future endodontic outcomes. Track 2 synthesized clinical decision domains relevant to NS-ReTx-versus-EMS selection. Risk of bias and applicability were assessed with PROBAST+AI; reporting was evaluated using TRIPOD+AI and CLAIM where applicable. Because of heterogeneity and limited evidence across outcomes, predictors, and performance metrics, findings were synthesized narratively.

RESULTS

Six AI prognostic studies met Track 1 eligibility, and 11 clinical studies informed Track 2 decision domains. Included models used radiographic, structured-variable, or hybrid approaches. All models used internal validation only; calibration, external validation, and clinical utility analyses were absent. No model generated treatment-conditional predictions comparing NS-ReTx and EMS within the same clinical scenario. Restorability, technical retrievability, and surgical-access variables were inconsistently represented.

CONCLUSIONS

Current AI prognostic models in endodontics remain early-stage single-pathway prediction tools and are not yet structured for comparative NS-ReTx-versus-EMS decision support.