Categories
Change Password!
Reset Password!
Full pulpotomy has become an increasingly accepted treatment option for permanent molars with symptomatic irreversible pulpitis, but predicting clinical outcomes remains challenging.
In pulpitis patients, machine learning models predict full pulpotomy success with moderate accuracy, with bleeding time emerging as the strongest clinical predictor.
Full pulpotomy has become an increasingly accepted treatment option for permanent molars with symptomatic irreversible pulpitis, but predicting clinical outcomes remains challenging. Researchers investigated whether machine learning (ML) models could accurately predict treatment success, postoperative pain intensity, and analgesic use using routinely collected clinical variables, with the aim of supporting individualized treatment planning and risk stratification.
Researchers conducted a retrospective multicohort study involving 214 permanent molars with symptomatic irreversible pulpitis and a minimum follow-up of 2 years after full pulpotomy. Multiple ML algorithms, encompassing logistic regression, support vector machines, and random forest, were developed for classification tasks. Ridge regression was utilized to forecast postoperative pain. Model performance was checked via the area under the receiver operating characteristic curve (AUC), accuracy, F1 score, and Brier score for calibration. Internal validation was performed through a training-test split and cross-validation.
Machine learning models illustrated the strongest predictive performance for treatment success after full pulpotomy. Classification models achieved moderate discriminative ability, with area under the receiver operating characteristic curve (AUC) values ranging from 0.68 to 0.75, while logistic regression performed best (AUC = 0.75). Bleeding time emerged as the most influential predictor of successful treatment outcomes.
In contrast, prediction of analgesic use was limited, with an F1 score of 0.19, likely reflecting class imbalance. Prediction of postoperative pain also showed low explanatory power, with a coefficient of determination (R²) of 0.21. Calibration analysis using Brier scores demonstrated moderate agreement between predicted and observed outcomes.
Machine learning-based prediction was most effective for identifying the likelihood of full pulpotomy success, while predictions of postoperative pain and analgesic requirements remained unreliable. The findings underscored the prognostic value of bleeding time and suggested that machine learning could complement, rather than replace, clinical judgment in treatment planning.
Journal of Endodontics
Machine Learning–Based Prediction of Treatment Success, Postoperative Pain, and Analgesic Use after Full Pulpotomy in Permanent Molars with Symptomatic Irreversible Pulpitis
Ecem Karakoyunlu et al.
Comments (0)