Objective:To evaluate the impact of deep learning reconstruction (DLR) on the image quality of carotid CT angiography (CTA), and to explore the visibility of black blood techniques for carotid atherosclerotic plaques and its ability to evaluate the degree of luminal stenosis.Methods:This was a cross-sectional study. A total of 122 carotid atherosclerotic plaques were retrospectively analyzed from 77 patients with soft plaques in the carotid artery CTA of Affiliated Zhongshan Hospital of Dalian University from March 2024 to March 2025. The CTA images were reconstructed using the hybrid iterative reconstruction (HIR) algorithm (HIR group) and DLR algorithm (DLR group). Objective image quality was evaluated by measuring the CT values and image noise (SD) of the carotid, sternocleidomastoid muscle and neck subcutaneous fat, then calculating the signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR). Subjective image quality was assessed using a 5-point scale. The reconstruction algorithm with better image quality was selected for black blood image reconstruction. In CTA and black blood images, the CT values and SD of soft plaques, sternocleidomastoid muscle and neck subcutaneous fat were measured, SNR and CNR were calculated, and visibility of plaques were scored using a 5-point scale. The objective indicators between two groups were compared using paired sample t-test or Wilcoxon signed rank test, and the subjective scores were compared using Wilcoxon signed rank test, the degree of luminal stenosis was compared using χ2 test, and the consistency evaluation was evaluated using weighted Kappa test. Results:In the DLR group, the CT values and SD of the aortic arch, brachiocephalic trunk, left common carotid artery origin, left subclavian artery origin, mid-segment of bilateral common carotid arteries, carotid bifurcation, internal carotid artery origin, and siphon were lower than those in the HIR group (all P<0.001). The SNR and CNR were higher than those in the HIR group (all P<0.001). The image quality score of CTA images in the DLR group was higher than that in the HIR group ( Z=7.44, P<0.001). Black blood images were reconstructed from DLR CTA images with higher image quality. The CT values, SNR, and CNR of soft plaques in the black blood images were higher than those in the DLR CTA images, and the SD was lower than that in the DLR CTA images (all P<0.05). Subjective evaluation showed that the visibility score of soft plaques in black blood images was higher than in DLR CTA images ( Z=-8.92, P<0.001). In DLR CTA images, there were 60 mildly narrowed lumens, 38 moderately narrowed lumens, and 24 severely narrowed lumens. In the black blood images, there were 54 mildly narrowed lumens, 39 moderately narrowed lumens, and 29 severely narrowed lumens. There was no statistically significant difference in the diagnosis of luminal stenosis between black blood images and DLR CTA images ( χ2=0.80, P=0.670), and the diagnostic consistency was good ( Kappa=0.893, P<0.001). Conclusion:DLR algorithm can significantly improve the image quality and reduce image noise. DLR algorithm combined with the black blood technique can significantly enhance the visualization of soft plaques.

Objective:To evaluate the impact of deep learning reconstruction (DLR) on the image quality of carotid CT angiography (CTA), and to explore the visibility of black blood techniques for carotid atherosclerotic plaques and its ability to evaluate the degree of luminal stenosis.Methods:This was a cross-sectional study. A total of 122 carotid atherosclerotic plaques were retrospectively analyzed from 77 patients with soft plaques in the carotid artery CTA of Affiliated Zhongshan Hospital of Dalian University from March 2024 to March 2025. The CTA images were reconstructed using the hybrid iterative reconstruction (HIR) algorithm (HIR group) and DLR algorithm (DLR group). Objective image quality was evaluated by measuring the CT values and image noise (SD) of the carotid, sternocleidomastoid muscle and neck subcutaneous fat, then calculating the signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR). Subjective image quality was assessed using a 5-point scale. The reconstruction algorithm with better image quality was selected for black blood image reconstruction. In CTA and black blood images, the CT values and SD of soft plaques, sternocleidomastoid muscle and neck subcutaneous fat were measured, SNR and CNR were calculated, and visibility of plaques were scored using a 5-point scale. The objective indicators between two groups were compared using paired sample t-test or Wilcoxon signed rank test, and the subjective scores were compared using Wilcoxon signed rank test, the degree of luminal stenosis was compared using χ2 test, and the consistency evaluation was evaluated using weighted Kappa test. Results:In the DLR group, the CT values and SD of the aortic arch, brachiocephalic trunk, left common carotid artery origin, left subclavian artery origin, mid-segment of bilateral common carotid arteries, carotid bifurcation, internal carotid artery origin, and siphon were lower than those in the HIR group (all P<0.001). The SNR and CNR were higher than those in the HIR group (all P<0.001). The image quality score of CTA images in the DLR group was higher than that in the HIR group ( Z=7.44, P<0.001). Black blood images were reconstructed from DLR CTA images with higher image quality. The CT values, SNR, and CNR of soft plaques in the black blood images were higher than those in the DLR CTA images, and the SD was lower than that in the DLR CTA images (all P<0.05). Subjective evaluation showed that the visibility score of soft plaques in black blood images was higher than in DLR CTA images ( Z=-8.92, P<0.001). In DLR CTA images, there were 60 mildly narrowed lumens, 38 moderately narrowed lumens, and 24 severely narrowed lumens. In the black blood images, there were 54 mildly narrowed lumens, 39 moderately narrowed lumens, and 29 severely narrowed lumens. There was no statistically significant difference in the diagnosis of luminal stenosis between black blood images and DLR CTA images ( χ2=0.80, P=0.670), and the diagnostic consistency was good ( Kappa=0.893, P<0.001). Conclusion:DLR algorithm can significantly improve the image quality and reduce image noise. DLR algorithm combined with the black blood technique can significantly enhance the visualization of soft plaques.