Workflow-level implant-position reproducibility in complete-arch scans under scanner-integrated artificial intelligence-based filtering and wireless scanner-to-cradle distance conditions: a device-specific in vitro study

Abstract

This device-specific in vitro study compared workflow-level implant-position reproducibility under scanner-integrated artificial intelligence (AI)-based filtering and wired/wireless scanner-to-cradle conditions. A maxillary edentulous model with six implants was scanned using one Aoralscan Elite unit, one Aoralscan Elite Wireless unit, and one host computer. AI-based filtering enabled and disabled conditions were defined as AI-ON and AI-OFF, respectively. Five scans were acquired for each of the eight AI × communication-condition combinations: wired reference and Wireless 0.5, 2.5, and 5.0 m (40 datasets). After computer-aided design library matching and virtual-abutment replacement, datasets were registered using three reference bodies. Concordance was the percentage of the virtual-abutment surface area within a |deviation| of ≤ 50 μm of the reference. With AI-OFF, the median concordance was 53.1% at 2.5 m and 50.0% at 5.0 m, compared to 97.7% and 98.7%, respectively, with AI-ON. Exploratory two-way analysis of variance showed an AI × communication-condition interaction [F(3, 32) = 14.92, p  < 0.001, partial η² = 0.583], supported by residual-permutation analysis. Holm-corrected exact comparisons showed higher concordance with AI-ON than with AI-OFF in the wired, 2.5-m, and 5.0-m conditions. Within this device-specific workflow, AI-ON was associated with high concordance across the tested conditions. Because only one unit of each scanner and one host computer were evaluated and because wireless quality was not measured directly, the findings do not establish a general AI or distance effect and should not be extrapolated to prosthesis fit or clinical performance.