Artificial Intelligence Benchmarking of Sparse-View Dental Cone-Beam Computed Tomography Reconstruction Under Unmeasured Patient Motion During Low-Dose Scanning: A Healthcare Image-Quality Robustness Evaluation

Sparse-view dental cone-beam computed tomography (CBCT) reconstruction with deep learning promises dose reduction, but clinical deployment is threatened by unmeasured patient motion that corrupts projection geometry in ways no benchmark currently captures. The design was evaluated across 2000 replications per cell over cluster counts of 8, 16, 24 and 32 and effect sizes of 3, 5 and 8 percentage points, under a declared variance structure with between-cluster SD of 6.0, residual SD of 4.5 and intraclass correlation of 0.64. Power analysis indicated that 54 clusters are required to detect a 5-percentage-point effect at 80% power. The framework distinguishes motion-agnostic, motion-aware and motion-robust benchmarking regimes, and shows that the BACI estimator's operating characteristics degrade sharply below 24 clusters. The contribution is a design-level account of what a defensible motion-robustness benchmark must measure, and a quantified statement of the sample sizes at which such benchmarks can distinguish reconstruction quality from noise.