深度学习重建 AiCE 技术在腹主动脉瘤腔内修复术后 CTA 中的应用.

Objective To study the application value of advanced intelligent clear-IQ engine (AiCE) of deep learning reconstruction (DLR) technology in CT angiography (CTA) after endovascular repair of abdominal aortic aneurysm (EVAR). Methods CTA images of 26 patients were reconstructed by hybrid iterative reconstruction (HIR) and DLR-AiCE methods. The quality of the overall images, liver and kidneys, vascular lumen within the stent and thrombus outside the stent at the level of maximum artifact was scored subjectively. The average of the four subjective scores was taken as the final score and the consistency of the score was analyzed. CT values and standard deviations of the liver, kidneys, aorta, and stent were calculated and the artifact index (AI), signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR) were compared. Results The subjective image quality scores were higher (all P＜0.001) in DLR-AiCE than that in HIR groups. There was a high agreement in subjective ratings between the two observers (ICC=0. 927, 95％CI:0. 905 to 0. 944, P＜0.001). Except for the minimum artifact within the right iliac stent and at the hepatic level, the AI value in the DLR-AiCE group was significantly (all P＜0. 05) lower than that in the HIR group. SNRs in DLR-AiCE group were significantly (all P＜0. 05) greater than that in the HIR group. Except for the maximum artifact in the right iliac stent and the minimum artifacts in the bilateral iliac stents, the CNRs in the DLR-AiCE group were significantly (all P＜0. 05) greater than that in the HIR group. Conclusion DLR-AiCE reduces stent artifacts and improves image quality of peripheral organs and aneurysm thrombus. DLR-AiCE is more conducive than HIR for CTA evaluation after EVAR [ABSTRACT FROM AUTHOR]