Prediction of Apical Extent Using Ensemble Machine Learning Technique in the Root Canal through Biomechanical Preparation: In-vitro Study

Abstract

This work aims to evaluate the dimensions of the apical extent after preflaring with the primary treatment and retreatment on human extracted teeth during endodontic treatment with the help of an ensemble machine learning model. The endodontic file ensures this procedure. It is a medical instrument utilized to eliminate the debris and smear layer as a pulp from the root canal during root canal treatment (RCT). Inadequate biomechanical RCT preparation frequently leads to post-operative apical periodontitis. This results in severe gum inflammation that harms the soft tissues, if left untreated, may harm the bones of the root canals supporting teeth. Therefore, to obtain the proper RCT instrumentation and endodontic treatment, the dimension of the apical extent has been analyzed using a machine learning model in this work. For this study, digital intraoral radiographic images have been recorded with the help of the Kodak Carestream Dental RVG sensor (RVG 5200). The RVG sensor is directly coupled with the CS imaging software (Carestream Dental LLC, NY) to acquire radiographs. Furthermore, the recorded images have been used to measure the dimensions of apical length. The machine learning ensemble classifiers are used in this study to classify the apical condition, such as apical extent, beyond the apical, and up to apical or perfectly RCT. The ensemble bagged, boosted, and RUSboosted trees classifiers are used in this analysis. The maximum accuracy obtained through the ensemble bagged trees model is 94.2 %, the highest among the models. The machine learning approaches can improve the treatment practice, improve RCT results, and provide a suitable decision support system.