Noise-Matched Blending Level Selection for 1024-Matrix CT Images Using Hybrid-Iterative Reconstruction: Comparison With 512-Matrix Images

OBJECTIVE: To identify the noise-matched hybrid-iterative reconstruction (IR) blending level for 1024-matrix images with noise equivalent to that of 512-matrix images while preserving comparable image quality. METHODS: In the phantom study, we identified the noise-matched hybrid-IR blending level for 1024-matrix images that achieved a noise level equivalent to 512-matrix images reconstructed with adaptive statistical iterative reconstruction-Veo (ASiR-V) 30%. This retrospective study included 63 patients (40 women; median age: 74y) who had undergone single-energy pancreatic protocol computed tomography (CT) from August 2023 to June 2024. For each patient, 3 image sets at the pancreatic phase were reconstructed: 512-matrix with ASiR-V 30% (group 1), 1024-matrix with ASiR-V 30% (group 2), and 1024-matrix with the noise-matched ASiR-V blending level (group 3). Background noise was measured as the SD of CT attenuation in the abdominal subcutaneous fat. Edge-rise slope (ERS) at the pancreas-fat interface was used to quantify edge sharpness. Two radiologists assessed overall image quality using a 5-point scale. RESULTS: The phantom study identified 60% as the noise-matched blending level for group 3. Background noise was significantly lower in the order of groups 3, 1, and 2, respectively (P < 0.001). ERS value was significantly higher in the order of groups 2, 3, and 1, respectively (P < 0.001). Overall image quality was significantly superior in the order of groups 3, 1, and 2, respectively (P < 0.001 for both radiologists). CONCLUSIONS: Increasing the matrix size leads to greater image noise with conventional reconstruction; however, applying a noise-matched hybrid-IR blending level allowed 1024-matrix images to maintain image quality while improving edge sharpness.