Cardiac-Specific AI Reconstruction Versus Hybrid Iterative Reconstruction for Emergency Coronary CT Angiography: A Paired Single-Center Study Including Triple-Rule-Out CT

Abstract

Background/Objectives: Optimizing coronary image quality is important in emergency computed tomography (CT), including coronary CT angiography (CCTA) and triple-rule-out CT angiography (TRO-CTA). We compared objective and subjective image quality between cardiac-specific artificial intelligence (AI) reconstruction and hybrid iterative reconstruction (HIR) using paired images from the same acquisition. Methods: This retrospective single-center study included 66 of 75 consecutively assessed emergency examinations: 28 standard CCTA and 38 TRO-CTA examinations acquired without pharmacologic heart-rate control. Both protocols used identical cardiac acquisition settings. AI and HIR images were reconstructed from the same acquisition with matched cardiac phases and motion-correction settings. Objective measurements were obtained at nine coronary locations, and two readers independently assessed overall image quality and perceived vessel-edge sharpness in unmasked paired comparisons. Results: AI reduced coronary noise from 45.8 ± 8.5 to 24.2 ± 7.5 Hounsfield units (HU; −47.3%), increased signal-to-noise ratio (SNR) from 9.0 ± 2.4 to 20.6 ± 5.6 (+129.6%), and increased contrast-to-noise ratio (CNR) from 14.8 ± 4.2 to 45.8 ± 16.4 (+209.7%; all Holm-adjusted p < 0.001). Readers 1 and 2 preferred AI for overall quality in 65/66 and 63/66 examinations and for sharpness in 64/66 and 63/66, respectively. Conclusions: Cardiac-specific AI reconstruction reduced image noise and increased SNR and CNR in emergency CCTA, including TRO-CTA, with readers generally favoring AI images for perceived quality and sharpness. Further studies should evaluate whether these image-quality improvements translate into better diagnostic performance.