Accuracy of manual and artificial intelligence‐based superimposition of cone‐beam computed tomography with digital scan data, utilizing an implant planning software: A randomized clinical study

Panagiotis Ntovas,Laurent Marchand,Matthew Finkelman,Marta Revilla‐León,Wael Att
DOI: https://doi.org/10.1111/clr.14313
2024-06-11
Clinical Oral Implants Research
Abstract:Objectives To investigate the accuracy of conventional and automatic artificial intelligence (AI)‐based registration of cone‐beam computed tomography (CBCT) with intraoral scans and to evaluate the impact of user's experience, restoration artifact, number of missing teeth, and free‐ended edentulous area. Materials and Methods Three initial registrations were performed for each of the 150 randomly selected patients, in an implant planning software: one from an experienced user, one from an inexperienced operator, and one from a randomly selected post‐graduate student of implant dentistry. Six more registrations were performed for each dataset by the experienced clinician: implementing a manual or an automatic refinement, selecting 3 small or 3 large in‐diameter surface areas and using multiple small or multiple large in‐diameter surface areas. Finally, an automatic AI‐driven registration was performed, using the AI tools that were integrated into the utilized implant planning software. The accuracy between each type of registration was measured using linear measurements between anatomical landmarks in metrology software. Results Fully automatic‐based AI registration was not significantly different from the conventional methods tested for patients without restorations. In the presence of multiple restoration artifacts, user's experience was important for an accurate registration. Registrations' accuracy was affected by the number of free‐ended edentulous areas, but not by the absolute number of missing teeth (p
engineering, biomedical,dentistry, oral surgery & medicine
What problem does this paper attempt to address?