Leveraging the Potential of Ultra-High Resolution Photon Counting Detector CT Angiography of the Cerebral Vasculature by Using Low kVp and Sharp Kernels.

Abstract

Objectives: To assess quality improvements in ultra-high resolution (UHR) photon counting detector (PCD) CT angiography (CTA) of the cerebral vasculature using image acquisition at lower kVp and image reconstruction with sharper kernels.

Materials and methods: All cerebral CTA imaging on PCD-CT performed over a period of 25 months was evaluated retrospectively. Records were excluded in case of protocol deviations, severe motion artifacts, or cerebral circulatory arrest. Using UHR resolution data acquisition (120×0.2 mm collimation) and reconstruction of polyenergetic (T3D) images at 0.2 mm, 3 subsequent protocols were evaluated: (1) 140 kVp/medium-sharp kernel (Hv40); (2) 90 kVp/Hv40; and (3) 90 kVp/very sharp kernel (Hv72). Virtual monoenergetic 0.4 mm reconstructions were derived at 55 keV using a medium-sharp kernel (Hv40) in protocols 1 and 2 and a sharp kernel (Hv60) in protocol 3. Vessel attenuation, signal-to-noise ratios (SNRs), and contrast-to-noise ratios (CNRs) were derived at 4 locations in the anterior circulation. Vessel sharpness was quantified using the edge rise distance and edge rise slope. Subjective assessments of image noise, vessel attenuation, and vessel sharpness were performed by 2 readers on a 5-point Likert scale.

Results: Out of 154 screened patient records, 141 were included. Vessel attenuation, SNR, and CNR improved with image acquisition at 90 kVp compared with 140 kVp for both 0.2 mm T3D images and, to a lesser extent, for 0.4 mm virtual monoenergetic images (VMI) in the protocols using a medium-sharp reconstruction kernel. In the 0.2 mm T3D images, SNR and CNR decreased when applying very sharp kernels at 90 kVp due to an increase in noise. However, SNR and CNR remained stable in small-caliber vessels, whereas both parameters decreased with medium-sharp kernels. Vessel sharpness was markedly improved in the 90 kVp/very sharp kernel protocol. Subjective assessment of image quality also favored the 90 kVp/very sharp kernel protocol. In the 0.4 mm VMI, similar improvements in quantitative and qualitative image quality were observed with image acquisition at 90 kVp (compared with 140 kVp) and image reconstruction using a sharp kernel (compared with a medium-sharp kernel).

Conclusions: Image quality of UHR PCD-CTA of the cerebral vasculature is improved with image acquisition at 90 kVp and image reconstruction with (very) sharp kernels.