70 kVp联合深度学习算法改进儿童腹部双低CTA图像质量的研究

Improving image quality of double-low abdominal CT angiography in children by using 70 kVp and deep learning image reconstruction

目的评价70 kVp结合深度学习图像重组(DLIR)算法在低辐射低对比剂用量情况下,是否可以提升0.625 mm薄层小儿腹部CTA图像质量。方法观察组选取连续的37例[7个月~14岁,平均(6.87±3.11)岁]腹部增强CTA检查,检查使用低辐射剂量,低对比剂检查方案,扫描采用70 kVp,对比剂用量0.8~1.2 ml/kg。采用0.625 mm层厚的高强度DLIR(DL-H)重组图像。对照组选取连续的34例儿童[年龄1~15岁,平均(6.34±3.11)岁],扫描采用常规放射剂量及对比剂用,采用100 kVp,对比剂用量1.0~1.6 ml/kg,采用50%自适应统计迭代重组法(ASIR-V)重组为0.625 mm层厚图像。两位放射科医生以5分标准(5分:优秀;4分:良好;3分:可测量;2分:可观察;1分:不合格)分别评估图像噪声、血管边缘清晰度和血管对比度。客观评价测量腹主动脉和竖脊肌的CT值和噪声值,并计算对比噪声比。采用配对t检验比较两组图像间的差异。结果观察组与对照组CT剂量指数无统计学差异[(1.47±0.26)mGy比(1.55±0.48)mGy,P>0.05]。观察组的对比剂用量为(28.19±13.47)ml,较对照组的(37.25±10.62)ml低24.32%(P<0.05)。肌肉和主动脉的图像噪声为14.50±3.54和20.48±5.74,显著低于对照组的19.01±4.57和24.90±3.53(P均<0.05)。观察组CNR为19.76±4.36,明显高于对照组的12.66±3.38(P<0.05)。主观评价DL-H图像符合诊断要求,噪声小于50%ASIR-V图像,整体图像质量优于50%ASIR-V。结论采用70 kVp结合DLIR算法在儿童低辐射剂量和对比剂量下能获得良好的0.625 mm薄层腹部CTA图像。

Objective To improve image quality of 0.625 mm thin slice pediatric abdominal CT angiography(CTA)with low radiation and contrast medium(CM)doses using 70 kVp and deep learning image reconstruction(DLIR)algorithm.Methods In the study group,37 children[age range:7 months-14 years,average:(6.87±3.11)years]underwent low radiation dose abdominal CTA with 70 kVp and low CM dose of 0.8-1.2 ml/kg.Images were reconstructed with DLIR at a high setting(DL-H)and slice thickness of 0.625 mm.In the control group,34 children[age range:1-15 years,average:(6.34±3.11)years]underwent routine dose abdominal CTA with 100 kVp and CM dose of 1.0-1.6 ml/kg.Images were reconstructed at slice thickness of 0.625 mm with 50%adaptive statistical iterative reconstruction-V(ASIR-V).Two radiologists evaluated all images in consensus for image noise,vessel margin and vessel contrast on a 5-point scale(5:excellent;4:good;3:measurable and acceptable;2:detectable;1:not acceptable).CT attenuation number and standard deviation of aorta and back muscle on the same image slice were measured and compared using paired-t test.Results The volume CT dose index value[(1.47±0.26)mGy]in the study group was not significantly different(P>0.05)from that of the control group[(1.55±0.48)mGy]whereas the 24.32%decrease in CM dose[(28.19±13.47)ml]of the study group was significant(P<0.05)compared to that of control[(37.25±10.62)ml].The image noise of muscle[(14.50±3.54)HU]and aorta[(20.48±5.74)HU]in the study group was significantly lower(both P<0.05)than that of control[(19.01±4.57)HU,(24.90±3.53)HU].The study group had significantly higher(P<0.05)contrast-to-noise ratio(19.76±4.36)than the control group(12.66±3.38).DL-H image quality of artery met diagnostic requirements while the 50%ASIR-V images were deemed to be too noisy.The overall image quality of DL-H was better than that of 50%ASIR-V.Conclusion It is feasible to improve image quality of 0.625 mm thin slice abdominal CTA with low radiation and CM doses in children using 70 kVp and DLIR algorithm.