Deep-learning CT reconstruction improves image quality but not diagnostic accuracy for sacral fragility fractures: MRI remains essential.

PURPOSE: To determine whether modern deep-learning CT reconstruction improves trauma-relevant diagnostic performance for sacral fragility fractures (SFF) and whether optimized CT can reduce the need for MRI in routine clinical practice. METHODS: In this retrospective paired-reader study, 171 pelvic CT examinations with corresponding MRI (reference standard) were analyzed (fracture prevalence 63.2%). Two blinded readers independently assessed each dataset reconstructed using standard iterative reconstruction and deep-learning reconstruction (DLR). Binary fracture detection and qualitative image quality were recorded. Diagnostic performance was compared using paired statistical testing. RESULTS: DLR significantly improved perceived image quality and diagnostic confidence across all evaluated domains (p < 0.001). However, no significant differences were observed in sensitivity, specificity, or overall diagnostic accuracy between reconstruction techniques for either reader (all p > 0.65). Diagnostic accuracy exceeded 87% across all conditions. Missed fractures on CT were primarily associated with nondisplaced or marrow-based injuries detectable on MRI. CONCLUSION: Although advanced CT reconstruction improves image quality and reader confidence, it does not significantly enhance diagnostic accuracy for sacral fragility fractures. Even with optimized reconstruction techniques, CT cannot reliably exclude SFF. MRI therefore remains essential in patients with persistent clinical suspicion despite negative or equivocal CT findings. LEVEL OF EVIDENCE: Level III.