Objective Accurate determination of root canal working length is critical for successful endodontic treatment. Traditional methods, such as radiographic imaging and apex locators, are subject to operator variability and technical limitations. This study evaluated the accuracy of artificial intelligence (AI)-assisted radiographic measurement compared with manual methods.Materials and Methods Fifty high-quality digital radiographs of extracted single-rooted teeth were analyzed. Tooth length was measured manually using a Boley Gauge, which served as the reference standard. AI-based measurements were generated using a multimodal large language model (ChatGPT) through standardized prompting. Agreement between methods was assessed using Bland-Altman analysis and intraclass correlation coefficient (ICC). Measurement time was also recorded.Results Bland-Altman analysis demonstrated close agreement between AI and manual measurements, with a small mean difference and slight tendency toward underestimation. The 95% limits of agreement indicated occasional clinically relevant discrepancies, including outliers up to ±2 mm. The ICC was 0.905 (95% CI not assessed), indicating excellent reliability. AI analysis required less time per image compared with manual measurement.Conclusion AI-assisted radiographic measurement shows strong agreement with manual methods under controlled conditions. However, given the presence of clinically significant outliers and the ex vivo design, these findings should be considered preliminary. Further validation in clinical settings and comparison with established methods such as apex locators are needed.KNOWLEDGE TRANSFER STATEMENTAI-assisted radiographic measurement may help dentists estimate working length more quickly and consistently during root canal treatment. While results closely matched manual measurements, occasional errors occurred. AI should be used to support, not replace, clinical judgment and established methods. Further clinical research is needed before routine use in endodontic practice.
Carlson et al. (Wed,) studied this question.