Developing an AI-powered tool for radiographic feedback on working length determination in pre-clinical endodontic training

Abstract

Background: Establishing an accurate working length is a critical step in root canal treatment and directly influences clinical success. As artificial intelligence increasingly integrates into medical education, applying it to enhance endodontic training has become increasingly important. Aims: This study aimed to develop a machine learning–based tool that provides prompt, personalized, constructive feedback on radiographic working length determination in a pre-clinical setting and to evaluate its usability among dental students. Methods: A newly labeled dataset of 3,000 radiographic images was created and categorized into optimal, over-extended, and under-extended working lengths. This dataset was balanced and split into 80%, 10%, and 10% for training, validation, and testing, respectively. Twenty-two convolutional neural network models were developed, trained, and evaluated using five diagnostic metrics (accuracy, F1-score, precision, recall, and testing time). The best-performing model was integrated into a web-based platform and piloted with 30 pre-clinical dental students who provided usability feedback via a Likert-scale questionnaire. The study hypothesized that students would rate the tool as usable and educationally supportive. Results: The custom-developed deep CNN achieved 97%–99% accuracy, 95%–98% F1-score, 94%–99% precision, and a recall rate of 96%–98%, with an average testing time of 0.54 s. Students rated the proposed system positively across clarity, ease of use, and learning support, with median usability scores of 5.0 across all items and interquartile ranges of 4–5 to 5–5. Conclusion: The AI-powered feedback system demonstrated high accuracy with strong user acceptance. By delivering instant, constructive feedback on working length determination, it supports effective learning and skill refinement in endodontic education. It is also beneficial in classrooms with large student populations. Future work will expand the dataset and integrate additional stages of root canal training into a unified AI-based educational platform.