Artificial Intelligence Benchmarking of Radiograph Pneumonia Classifiers Under Unmeasured Portable-Device Scatter Correction Drift: A Healthcare Calibration-Transfer Evaluation

Portable chest radiography introduces scatter-correction drift that is rarely recorded in the metadata accompanying benchmark datasets, so pneumonia classifiers trained on fixed-room images may be evaluated under a calibration regime that does not match deployment. the present analysis treats that mismatch as a design problem rather than a data problem and characterises the operating characteristics of a before-after-control-impact (BACI) benchmarking design for detecting calibration transfer failure. No observational data are used or claimed. Under a declared variance structure with between-cluster SD = 6.0, residual SD = 4.5 and intraclass correlation = 0.640, power was estimated across 2000 replications per design cell for cluster counts of 8, 16, 24 and 32 and effect sizes of 3, 5 and 8 percentage points. The design required 54 clusters to detect a 5-percentage-point calibration-transfer effect at 80% power. The analysis formalises calibration transfer as a contrast between treated and untreated acquisition strata, derives the BACI estimator and its variance, and reports power curves, minimum detectable effects and coefficient behaviour. The contribution is a design-level account of how unmeasured scatter-correction drift propagates into classifier benchmarking, together with the cluster counts at which a BACI evaluation can distinguish drift-induced miscalibration from sampling noise.