Influence of an artificial intelligence-based application on the accuracy of complete arch implant scans recorded by using an intraoral scanner.

Statement Of Problem: Artificial intelligence (AI)-based applications have been integrated into intraoral scanner (IOS) programs aiming to assist the digitizing procedure, including the automatic alignment of the computer-aided design (CAD) file of the selected implant scan body (ISB). However, the impact of this AI-based application on the accuracy of complete arch implant scans remains unknown.

Purpose: The purpose of this in vitro study was to assess the impact of an AI-based application on the accuracy of complete arch implant scans recorded by using an IOS.

Material And Methods: A maxillary edentulous stone cast with 6 implant abutment analogs (MultiUnit Abutment Plus Replica) was obtained. Three groups were created based on the technique used to record complete arch implant scans by using an IOS (i700): prescan with the posterior ISB scan activating (Pre-AI group) or disabling the AI-based application (Pre-NoAI group) or obtaining a complete arch implant scan without prescan (NoPre group) (n=10). A complete arch scan involving the implant abutments was obtained (prescan) and duplicated 20 times. Then, an ISB (Elos Accurate IO 2C-A) was hand tightened into each implant abutment analog. In the Pre-AI group, for each duplicated prescan, each ISB was scanned until each scanned ISB was automatically aligned with the CAD file by the AI-based program. In the Pre-NoAI group, for each duplicated prescan, the ISBs were digitized to obtain the experimental complete arch implant scans. In the NoPre group, a complete arch implant scan was obtained. Lastly, a laboratory scan (reference scan) was obtained (T710). For each scan, an implant-supported bar was designed by using a program (DentalCAD). The bar designs were imported into another program (Geomagic), and linear and angular measurements among the implants were made. The measurements obtained in the control scan were used as a reference to measure the scanning distortion of each specimen. A 1-way Welch ANOVA test followed by the post hoc pairwise multiple comparison Tukey test was used to analyze trueness (α=.05). The Levene test was used to analyze precision (α=.05).

Results: Significant linear (P<.001) trueness discrepancies were found among the groups. The Pre-AI group obtained significantly worse linear trueness than the other 2 groups. Additionally, significant angular trueness (P<.001) and precision (P<.001) discrepancies were revealed among the groups. The Pre-AI group obtained significantly worse angular trueness and precision than the other groups tested.

Conclusions: The implant scanning technique impacted the 3-dimensional implant positions recorded by using the IOS tested.