Clinical Feasibility of an Artificial Intelligence-Assisted Digital Workflow for Full-Arch Implant Rehabilitation: A Multicenter Case Series

Abstract Fully digital workflows are increasingly used in implant prosthodontics; however, completely edentulous full-arch rehabilitations remain challenging because implant position transfer and passive framework fit are susceptible to cumulative scanning and registration errors. Clinical evidence on artificial intelligence (AI)-assisted complete-arch implant impression workflows remains limited. This multicenter case series aimed to describe the clinical feasibility of an AI-assisted digital workflow (SmartX) combined with extended scan bodies for all-on-X full-arch implant-supported rehabilitations. Ten completely edentulous patients rehabilitated with all-on-X implant-supported full-arch prostheses were included. Digital impressions were obtained using SmartFlag scan bodies and an AI-assisted SmartX workflow with an intraoral scanner. The primary outcome was clinically acceptable passive fit, assessed qualitatively by visual/tactile inspection, the one-screw (Sheffield) test, and a screw-resistance test, all recorded as dichotomous outcomes (acceptable/not acceptable). Secondary outcomes were implant and prosthesis survival and biological or technical complications during a minimum 6-month follow-up. A total of 54 implants were placed (mean: 5.4 implants per patient). All 10 cases demonstrated clinically acceptable passive fit according to the three qualitative clinical assessments. Because of the small sample size, the observed 100% positive outcome should be interpreted cautiously; the exact 95% confidence interval for 10/10 positive cases is wide. Progressive screw tightening showed an approximately 60-degree angular displacement from 5 to 20 Ncm in all cases; this observation was used only as an indirect clinical indicator of seating behavior and not as a quantitative measurement of misfit. No implant or prosthesis failures and no biological or technical complications were observed during the short follow-up period. Within the limitations of a small uncontrolled case series with qualitative outcomes and short follow-up, the investigated AI-assisted workflow appeared clinically feasible for full-arch implant-supported rehabilitation. These findings should be considered preliminary and require confirmation through controlled clinical studies with quantitative trueness/misfit measurements and longer follow-up.