THERE: Toward an easy and reliable tool for automatic cephalometric analysis

Background: Cephalometric analysis in orthodontics is a meticulous process requiring high precision in identifying anatomical landmarks on lateral cephalometric radiographs. Manual analysis by clinicians remains the standard, as even slight deviations in landmark placement can lead to incorrect diagnoses. Despite advancements in AI, automatically localizing these landmarks remains challenging, with few systems meeting the exacting standards of European orthodontic practice. Methods: We developed THERETransKey, a model designed for the automatic and accurate localization of anatomical landmarks in cephalometric analysis. To evaluate its practical utility, we integrated the model into THERE (auTomatic HElpeR for cEphalometry), an open-access, user-friendly AI-based tool for cephalometric tracing. Unique in its strict adherence to the guidelines of the European Board of Orthodontics, THERE provides immediate online access and continuously collects user data to improve its performance. The tool has been validated by users with different levels of experience in Orthodontics using a PSSUQ-based questionnaire. Results: THERETransKey has shown enhanced accuracy with respect the previous models underlying THERE, as confirmed by its integration into daily clinical workflows at the University of Ferrara. Moreover, user feedback collected through the administered questionnaire confirms THERE's improved usability. Finally, its regular use enables the generation of a clinician-validated dataset during everyday practice. Conclusions: By effectively supporting clinicians, THERE not only enhances cephalometric analysis but also contributes to building a robust dataset for future research. This initiative promotes a more generalized approach to automatic landmark detection across diverse radiographic equipment.