The application value of deep learning-based reconstruction combined with segmented contrast injection in "one-stop"...

"One-stop" cardio-cerebrovascular computed tomography angiography (C&C-CTA) has become a current mainstay of craniocervical and coronary arteries screening, but concerns remain regarding the radiation safety and contrast nephropathy issues associated with CTA examinations. This study aimed to investigate the feasibility of reducing radiation dose, contrast dose, and injection rate using C&C-CTA based on deep learning image reconstruction (DLIR) algorithms.

A total of 150 patients were prospectively enrolled and divided into 3 groups (A, B, and C) with 50 patients in each group. In group A, C&C-CTA was performed with a tube voltage of 80 kVp, a contrast injection protocol with segmented contrast injection, and high-strength DLIR. In groups B and C, craniocervical and coronary CTA were performed with 100 kVp and 50% adaptive statistical iterative reconstruction-V (ASIR-V), respectively. Radiation dose, contrast dose, injection rate, and subjective and objective image quality were quantified and compared.

Compared with group B, group A showed higher computed tomography (CT) Hounsfield unit (HU) values for the craniocervical arteries (all P<0.001). In contrast, when compared with group C, group A had lower CT HU values for the coronary artery roots (P=0.02), yet no statistically significant difference was observed in the values for other vessels (both P>0.05). The signal-to-noise ratios (SNRs) were not statistically different in vertebral artery, internal carotid artery (ICA) C4 segment (A vs. B), and right coronary artery (A vs. C) (all P>0.05), and other SNR were all statistically different and higher in group A (all P<0.05). The contrast-to-noise ratios (CNRs) in group A were all at a higher level, whereas the background noise levels were all at a lower level, and there were significant statistical differences (all P<0.001). The craniocervical arteries in group A were sharper than those in group B (both P<0.05), whereas the sharpness of the coronary arteries did not significantly differ from that in group C (P>0.05). The subjective scores in coronary and craniocervical arteries were all not statistically different (all P>0.05). Compared to the sum of B and C, the effective dose and contrast dose were reduced by 43.78% and 41.25% in group A, respectively.

Compared to scanning the coronary and craniocervical artery individually, C&C-CTA based on DLIR can reduce the radiation dose, contrast dose, and injection rate without compromising image quality.