Quantitative low-kVp CT angiography in carotid artery imaging

Shading and streaking artifacts brought by beam hardening and photon starvation is commonly observed in lowkVp CT angiography (CTA) for carotid artery imaging. Image quality is thus significantly degraded and a faithful observation of carotid artery is impeded. In this paper, we propose a novel quantitative image postprocessing scheme to eliminate these artifacts. In shading correction, we follow general knowledge of the relatively uniform CT number distribution in one tissue component, and continuous and low-frequency shading artifact distribution in projection domain. Coarse image segmentation is first applied to construct an ideal template image where each structure is filled with the same CT number of that specific tissue. By forward projecting the difference between uncorrected CT image and the ideal template, we estimate the continuous and lowfrequency shading error signal in projection domain using lowpass filtering. An error map is then reconstructed using standard filtered back-projection algorithm from the error signal and added to the original image to correct for the shading artifacts. To suppress the increased noise and streaking artifacts, we first perform a texture extraction to estimate the noise distribution in the image and then incorporate the estimated noise variation into a nonlocal filtering method. The proposed scheme is evaluated in carotid CTA scan using a dual-source CT at 80 kVp, in which the shading and streaking artifacts can be successfully corrected by reducing the CT number error from 246 HU to 63 HU and spatial non-uniformity by a factor of 2.2, respectively. In the volume rendering generated from the corrected CT images, the visualization of carotid artery is improved substantially and comparable to that generated from a CTA scan at 140-kVp. The proposed method is implemented directly on the CT image without access to the raw projection data and is thus attractive for clinical application in low-dose CTA.