Evaluating the Accuracy of Deep Learning Models and Dental Postgraduate Students in Measuring Working Length on Intraoral Periapical X-rays: An In vitro Study

Background: The integration of artificial intelligence in dentistry has seen remarkable advancements, especially in diagnostic imaging. This study evaluates and compares the accuracy of deep learning models with that of dental postgraduate students in determining working length on intraoral periapical radiographs. Materials and Methods: One hundred anonymized radiographs of single-rooted teeth with files at working length were obtained. The images were preprocessed and used to train a deep learning model. Five dental postgraduates visually estimated the working length after receiving training. Pixel counting in image processing software provided the gold standard measurement. Accuracy comparisons were performed using a t-test. Results: The deep learning model demonstrated significantly higher accuracy (85%) compared to human estimations (mean accuracy 75.4%). The t-test yielded P = 0.0374 (P < 0.05), rejecting the null hypothesis. Conclusion: Deep learning models show great potential in enhancing precision and reliability for working length determination in endodontics. With further refinement, these models can effectively complement human expertise in dental radiographic interpretation.