影像诊断与介入放射学

目的 研究深度学习重建算法(DLR)对低管电压、低对比剂用量的头颈CT血管造影(CTA)图像质量的影响。方法 前瞻性收集我院2022年4月—2023年11月行头颈CTA检查的患者66例，并随机分为两组。实验组采用80kVp管电压，注射对比剂碘帕醇(370 mg I/mL)28 mL，生理盐水40 mL，采用混合迭代重建(HIR)和DLR的低、中、高强度算法重建出4组图像，分别命名为L-HIR组、L-DLR_mild组、L-DLR_standard组、L-DLR_strong组。对照组采用100 kVp管电压，注射对比剂40 mL，生理盐水40 mL，采用HIR重建，命名为R-HIR组。测量并计算主动脉弓、颈内动脉、颈外动脉、椎动脉、基底动脉、大脑中动脉平均CT值、图像噪声(SD)、信噪比(SNR)和对比噪声比(CNR)等客观指标，并对5组图像进行5分制主观评分。根据数据是否满足正态分布，选择采用独立样本的t检验或者Mann-Whitney U检验来比较各组的图像质量。结果 3组L-DLR图像在主动脉弓、基底动脉、椎动脉、颈内动脉的CT值均大于R-HIR组(P<0.05),L-DLR_mild组在主动脉弓、颈内动脉层面的SD值与R-HIR组无显著差异(P>0.05)，其余L-DLR组的SD值均小于R-HIR(P<0.05),L-DLR_standard组和L-DLR_strong组的SNR值均大于R-HIR(P<0.05),L-DLR_strong组的CNR值均大于R-HIR(P<0.001)。5组图像的主观评分分别为：R-HIR(3.93±0.27)、L-HIR(2.72±0.82)、L-DLR_mild(3.34±0.82)、L-DLR_standard(3.52±0.6)、L-DLR_strong(3.61±0.79)，除L-HIR组外均符合诊断要求。实验组有效辐射剂量较对照组减少了65%[(0.41±0.08)mSv比(1.18±0.12)m Sv]。结论 采用80 kVp管电压和28 mL对比剂，结合深度学习重建算法可在客观上获得与100 kVp管电压和40 mL对比剂相当的头颈CTA图像质量。该方法减少辐射剂量和对比剂用量，提高头颈CTA检查的安全性。

Objective To investigate the effect of deep learning reconstruction (DLR) algorithm on the image quality of head and neck CT angiography (CTA) with low tube voltage and low contrast agent dosage.Methods A total of 66 patients who underwen head and neck CTA in our hospital from April 2022 to November 2023 were prospectively enrolled and randomly divided into two groups CTA of the experimental group was performed using 80 kVp tube voltage,28 mL of contrast agent,and 40 mL of saline.Images were reconstructed with hybrid iterative reconstruction (HIR),DLR at mild-,standard-and strong-strength producing four datasets of low-dose HIR (L-HIR),L-DLR_mild,L-DLR_standard,and L-DLR_strong.CTA of the control group was performed using 100 kVp tube voltage,40 mL of contrast agent,and 40 mL of saline with HIR producing regular-dose HIR dataset (R-HIR).Objective indexes such as aortic arch,interna carotid artery,external carotid artery,vertebral artery,basilar artery,middle cerebral artery,image noise (standard deviation,SD)signal-to-noise ratio (SNR),and contrast-to-noise ratio (CNR) were determined and the five groups of images were subjectively scored.The independent samples t-test or Mann-Whitney U test was used to compare the image quality between the two groups.Results The CT values of three L-DLR image groups in the aortic arch,basilar artery,vertebral artery,and internal carotid artery were all significantly higher than those of the R-HIR group (P<0.05).The L-DLR_mildgroup showed no significant difference at the aortic arch and internal carotid artery levels from that of the R-HIR group (P>0.05) whereas the SD values were significantly lower and SNR significantly higher in the L-DLR_standardand L-DLR_stronggroups than those of R-HIR group (P<0.05).The CNR of the L-DLR_stronggroup was significantly higher than that of the R-HIR group (P<0.001).The subjective scores of R-HIR (3.93±0.27),L-DLR_mild(3.34±0.82),L-DLR_standard(3.52±0.6)and L-DLR_strong(3.61±0.79) met diagnostic requirements whereas the L-HIR score (2.72±0.82) was nondiagnostic.The effective radiation dose in the experimental group[(0.41±0.08) mSv]was reduced by 65%compared to that of the control group[(1.18±0.12mSv)].Conclusion The image quality of head and neck CTA with 80 kVp and 28 mL of contrast agent combined with DLR algorithm is objectively equivalent to that with 100 kVp and 40 mL of contrast agent.The radiation dose and contrast agent dosage were significantly reduced,thus improving the safety of head and neck CTA.