Peter Melcher, Ralf Henkelmann, Salim Youssef, Anton Brehmer, Timm Denecke, Jeanette Henkelmann
Modeling postacquisition speckle variability during training improves the generalizability of OCT classifiers without increasing the inference cost, thereby enabling the more reliable deployment of artificial intelligence (AI)-assisted retinal screening across heterogeneous imaging systems.
Background/Objectives: Accurate patient positioning is essential for diagnostic quality in conventional knee radiography. Although established anatomical landmarks are routinely used to judge projection quality, assessment remains qualitative and observer-dependent. We therefore developed and validated the Radiographic Projection Quality Score for the Knee (RPQS-Knee), a standardized scoring system for objective projection quality assessment. Methods: The RPQS-Knee was derived from established radiographic positioning criteria, anatomical reference landmarks, and predefined geometrically based measurement thresholds. The final score was evaluated in 153 consecutive knee radiographic examinations, independently assessed by one radiologist and two orthopedic surgeons. Inter rater reliability was assessed using quadratically weighted Gwet's AC2, intra rater reliability using weighted Cohen's κ, and construct validity by correlation with separately assigned diagnostic adequacy ratings. Results: The RPQS-Knee comprises four projection-related criteria assessing fibular-head superimposition, tibial plateau alignment, and femoral-condyle congruency. Inter rater reliability was excellent for the total score (AC2 = 0.93; 95% CI, 0.92-0.95) and good to excellent for all individual items (AC2 = 0.76-0.85). Intra rater reliability ranged from κ = 0.59 to 0.81. RPQS-Knee subscores correlated strongly with diagnostic adequacy ratings (ρ = -0.69 and -0.78; both p < 0.001), providing evidence supporting construct validity. Conclusions: The RPQS-Knee translates established radiographic positioning criteria into a standardized, reproducible instrument for objective assessment of projection quality. Its excellent reliability and evidence supporting construct validity suggest that it may facilitate standardized quality assessment in clinical practice, quality assurance, and research.