EVALUATING THE ACCURACY OF MEASURING ENDODONTIC WORKING LENGTH FROM DIAGNOSTIC RADIOGRAPHS USING ARTIFICIAL INTELLIGENCE: AN IN VITRO STUDY

Successful root canal treatment largely depends on accurate working length determination, as errors in measurement may lead to incomplete cleaning or over-instrumentation of the canal. Conventional techniques such as radiographic assessment and electronic apex locators are widely used, but they may be influenced by operator experience, anatomical variations, and image interpretation errors. With recent advances in artificial intelligence (AI), deep learning models have shown promising applications in dental imaging and automated diagnosis. The present study was conducted to evaluate the accuracy of measuring endodontic working length from diagnostic radiographs using an AI-based approach. Diagnostic radiographs of single-rooted teeth were collected and processed through image enhancement and normalization techniques. A convolutional autoencoder model was used to identify and segment the root canal area, after which skeletonization and contour analysis were performed to calculate the canal length. The measurements obtained through AI were then compared with those measured using RVG software. The findings demonstrated a high level of agreement between the two methods, with accuracy values ranging from 96.93% to 99.5% and an average accuracy of approximately 98.39%. The AI-assisted system was able to provide consistent and reliable measurements while minimizing manual intervention. Within the limitations of this study, the proposed AI model appears to be a promising tool for improving the precision and reproducibility of working length determination in endodontic practice.