Predictive Machine learning Models for necessity Supplemental Anesthesia in Endodontic treatment

Abstract

Purpose: In cases of irreversible pulpitis, controlling intraoperative endodontic discomfort is extremely difficult, and patient satisfaction plays a big part in this. In order to forecast a diagnostic's and a treatment's outcome, machine learning (ML) has recently been implemented in the fields of medicine and dentistry. The goal of this work was to create machine learning (ML) models that could predict the need for further anesthesia. Methods: According to inclusion and exclusion criteria, this study included 128 individuals with endodontic discomfort. All patients underwent a clinical evaluation and endodontic diagnostic procedures. All interpretation results were entered into a prepared data sheet. All info were statistically evaluated using Github software version ydata-profilling vv4.1.2, configuration config.json, was employed to review the explanatory data for machine learning models for all examination and investigation aspects. By using Pearson correlation, chi 2, Random Forest, and LightGBM, the final feature importance was determined. 20% of the test set and 80% of the train set are observation sets used to build models. Logistic regression F1 and k-nearest neighbors (KNN) F1 were used to assess the performance of the ML model on the train and test sets. Results: For Machine learning models, 11 of the 20 features—such as pulp stone or calcification of the pulp space, pain duration, age, percussion, palpation, response persistent after EPT, dental history, curved root canal, pain persistent after a cold test, and pain severity during a cold test—were important. In logistic regression, F1 for the train set was 0.793, while for the test sets, it was 0.878. Regression using a logistic model had an accuracy of 0.81. KNN F1 for train was 0.781, while for test it was 0.829. The Machine learning model's k-nearest neighbors (KNN) F1 accuracy was 72.86. Conclusion: The trained machine learning models can predict if further anesthetic will be required during endodontic treatment based on the specific feature.