Objective: To assess noise reduction achievable with an iterative reconstruction algorithm.
Methods: 32 consecutive chest CT angiograms were reconstructed with regular filtered back projection (FBP) (Group 1) and an iterative reconstruction technique (IRIS) with 3 (Group 2a) and 5 (Group 2b) iterations.
Results: Objective image noise was significantly reduced in Group 2a and Group 2b compared with FBP (p < 0.0001). There was a significant reduction in the level of subjective image noise in Group 2a compared with Group 1 images (p < 0.003), further reinforced on Group 2b images (Group 2b vs Group 1; p < 0.0001) (Group 2b vs Group 2a; p = 0.0006). The overall image quality scores significantly improved on Group 2a images compared with Group 1 images (p = 0.0081) and on Group 2b images compared with Group 2a images (p < 0.0001). Comparative analysis of individual CT features of mild lung infiltration showed improved conspicuity of ground glass attenuation (p < 0.0001), ill-defined micronodules (p = 0.0351) and emphysematous lesions (p < 0.0001) on Group 2a images, further improved on Group 2b images for ground glass attenuation (p < 0.0001), and emphysematous lesions (p = 0.0087).
Conclusion: Compared with regular FBP, iterative reconstructions enable significant reduction of image noise without loss of diagnostic information, thus having the potential to decrease radiation dose during chest CT examinations.