Raw data-based iterative reconstruction in body CTA: evaluation of radiation dose saving potential

To evaluate prospectively, in patients undergoing body CTA, the radiation dose saving potential of raw data-based iterative reconstruction as compared to filtered back projection (FBP). Twenty-five patients underwent thoraco-abdominal CTA with 128-slice dual-source CT, operating both tubes at 120 kV. Full-dose (FD) images were reconstructed with FBP and were compared to half-dose (HD) images with FBP and HD-images with sinogram-affirmed iterative reconstruction (SAFIRE), both reconstructed using data from only one tube-detector-system. Image quality and sharpness of the aortic contour were assessed. Vessel attenuation and noise were measured, contrast-to-noise-ratio was calculated. Noise as image quality deteriorating artefact occurred in 24/25 (96%) HD-FBP but not in FD-FBP and HD-raw data-based iterative reconstruction datasets (p < 0.001). Other artefacts occurred with similar prevalence among the datasets. Sharpness of the aortic contour was higher for FD-FBP and HD-raw data-based iterative reconstruction as compared to HD-FBP (p < 0.001). Aortoiliac attenuation was similar among all datasets (p > 0.05). Lowest noise was found for HD-raw data-based iterative reconstruction (7.23HU), being 9.4% lower than that in FD-FBP (7.98HU, p < 0.05) and 30.8% lower than in HD-FBP images (10.44HU, p < 0.001). Contrast-to-noise-ratio was lower in HD-FBP (p < 0.001) and higher in HD-raw data-based iterative reconstruction (p < 0.001) as compared to FD-FBP. Intra-individual comparisons of image quality of body CTA suggest that raw data-based iterative reconstruction allows for dose reduction > 50% while maintaining image quality. Key Points aEuro cent Raw data-based iterative reconstruction reduces image noise and improves image quality as compared to filtered back projection aEuro cent At a similar radiation dose, raw data-based iterative reconstruction improves the sharpness of vessel contours aEuro cent In body CTA a dose reduction of > 50% might be possible when using raw data-based iterative reconstructions, while image quality can be maintained.