切换至 "中华医学电子期刊资源库"

中华腔镜泌尿外科杂志(电子版) ›› 2022, Vol. 16 ›› Issue (06) : 539 -545. doi: 10.3877/cma.j.issn.1674-3253.2022.06.011

临床研究

深度学习图像重建双低技术在CT尿路造影中的初步应用
张悦1, 张可1, 郭月飞1, 郭焯欣1, 孟占鳌1,()   
  1. 1. 510630 广州,中山大学附属第三医院放射科
  • 收稿日期:2022-05-19 出版日期:2022-12-01
  • 通信作者: 孟占鳌
  • 基金资助:
    广州市科技计划项目(202007030007)

Preliminary application of deep learning image reconstruction double-low technique in CT urography

Yue Zhang1, Ke Zhang1, Yuefei Guo1, Zhuoxin Guo1, Zhan'ao Meng1,()   

  1. 1. Department of Radiology, the Third Affiliated Hospital, Sun Yat-sen University, Guangzhou 510630, China
  • Received:2022-05-19 Published:2022-12-01
  • Corresponding author: Zhan'ao Meng
引用本文:

张悦, 张可, 郭月飞, 郭焯欣, 孟占鳌. 深度学习图像重建双低技术在CT尿路造影中的初步应用[J]. 中华腔镜泌尿外科杂志(电子版), 2022, 16(06): 539-545.

Yue Zhang, Ke Zhang, Yuefei Guo, Zhuoxin Guo, Zhan'ao Meng. Preliminary application of deep learning image reconstruction double-low technique in CT urography[J]. Chinese Journal of Endourology(Electronic Edition), 2022, 16(06): 539-545.

目的

探讨100 kV深度学习图像重建(DLIR)在CT尿路造影(CTU)中双低(低辐射剂量、低对比剂剂量)技术的可行性,并与传统的120 kV自适应统计迭代重建-V(ASIR-V)标准方案进行比较。

方法

收集60例行腹部CTU增强扫描的患者按扫描方案分为标准方案组(S组)和双低方案组(L组),记录L组和S组对比剂剂量、容积CT剂量指数和剂量长度乘积,计算有效剂量。对S组采用60%ASIR-V(S-AV60)重建,L组采用60%ASIR-V(L-AV60)、DLIR-M、DLIR-H重建。测量右侧肾盂、右侧肾脏实质、左侧肾盂、左侧肾脏实质、右侧输尿管、左侧输尿管、膀胱、腰大肌CT值和标准差(SD),计算信噪比(SNR)和对比噪声比(CNR)。由两名临床经验丰富的诊断医师对原始轴位影像、重建后容积再现(VR)图像和最大强度投影(MIP)图像进行双盲法主观评分。两组有效剂量(ED)比较采用t检验,CT、SD、SNR、CNR值等客观评价参数的分析采用单因素方差分析,主观评分采用Kruskal-Wallis检验,用Kappa检验分析2名放射科医师主观评分的一致性。

结果

L组的有效剂量和对比剂剂量分别比S组减少了29.1%(P<0.001)和32.1%(P<0.001),DLIR-H组的SD最低,SNR,CNR最大。四个因素的独立分数和图像最终评分,DLIR-H组图像评分最高(P<0.001)。两名放射科医师对4个CTU的主观图像质量评分亦有很好的一致性(Kappa=0.770,P<0.001)。

结论

在CT尿路造影(CTU)中,DLIR重建的双低技术可显著降低放射剂量(29.1%)、对比剂剂量(32.1%)。与60%ASIR-V标准方案相比,DLIR-H可以进一步改善图像质量,是较好的重建算法。

Objective

To explore the feasibility of 100 kV deep learning image reconstruction (DLIR) in CT urography (CTU) and compare them with the traditional 120kV adaptive statistical iterative reconstruction-V (ASIR-V) standard scheme.

Methods

According to the scanning scheme, 60 patients with abdominal CTU enhancement scan were divided into standard plan group (group S) and double low plan group (group L). The contrast dose, volume CT dose index and dose length product of group L and group S were recorded, and the effective dose was calculated. Group S was reconstructed with 60%ASIR-V (S-AV60), and group L was reconstructed with 60%ASIR-V (L-AV60), DLIR-M and DLIR-H. The CT values and standard deviation (SD) of right renal pelvis, right renal parenchyma, left renal pelvis, left renal parenchyma, right ureter, left ureter, bladder and psoas major muscle were measured, and the signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR) were calculated. The original axial images, reconstructed volume rendering (VR) images and maximum intensity projection (MIP) images were subjectively scored by two experienced clinicians. t-test was used to compare ED between the two groups. One-way ANOVA was used to analyze the objective evaluation parameters such as CT, SD, SNR and CNR. Kruskal-Wallis test was used to analyze the subjective score of the two radiologists. Kappa test was used to analyze the consistency of subjective scores of the two radiologists.

Results

Compared with group S, the effective dose and contrast dose of group L decreased by 29.1% (P<0.001) and 32.1% (P<0.001), respectively. The SD of DLIR-H group was the lowest and the SNR,CNR was the largest. The independent score and the final image score of the four factors were the highest in the DLIR-H group (P<0.001). The subjective image quality scores of 4 CTU were also consistent among 2 radiologists (Kappa=0.770, P<0.001).

Conclusions

In CT urography (CTU), the double-low technique of DLIR reconstruction can significantly reduce the radiation dose (29.1%) and contrast dose (32.1%). Compared with the 60%ASIR-V standard scheme, DLIR-H can further improve the image quality and is a better reconstruction algorithm.

表1 两组接受CTU增强检查患者一般数据和辐射剂量比较(±s
图3 不同组别CTU重建图像客观参数比较注:图3a、3b、3c、3d分别为S组与L组右侧肾盂,左侧肾盂,右侧输尿管,左侧输尿管,膀胱中CT、SD、SNR、CNR值直方图。图3a中CT值S组<L组(P<0.001),图3bSDL-DH<其他各组(P<0.001),图3c3dSNR值、CNRL-DH>其他各组(P<0.001),其中实线为L-DH与其他各组组间差异有统计学意义(P<0.001)
表2 不同组别CTU重建图像客观参数比较(±s
组别 例数 CT值(Hu) SD值
右肾盂 左肾盂 右输尿管 左输尿管 膀胱 右肾盂 左肾盂
S-AV60 30 983.06±328.53 1022.06±351.61 926.44±316.86 944.63±388.42 732.71±366.29 71.41±11.61 71.12±22.03
L-AV60 30 1285.53±527.87 1277.06±402.30 1291.45±490.11 1254.27±484.89 999.09±458.84 92.47±55.85 109.63±67.75
L-DM 30 1291.91±529.94 1289.36±401.57 1248.36±410.01 1251.14±482.09 1000.38±459.04 85.80±55.50 74.65±62.27
L-DH 30 1292.15±529.98 1289.41±401.67 1273.65±469.81 1251.44±481.84 1000.43±458.98 44.93±22.57 35.87±24.09
F 4.579 3.425 4.939 3.339 2.797 13.883 18.064
P 0.006 0.020 0.003 0.022 0.043 <0.001 <0.001
组别 SD值 SNR值
右输尿管 左输尿管 膀胱 右肾盂 左肾盂 右输尿管 左输尿管 膀胱
S-AV60 93.4±25.86 90.92±25.41 46.68±30.31 13.88±3.99 15.79±7.56 10.40±3.84 10.71±4.07 17.22±5.55
L-AV60 164.55±109.47 160.96±107.43 65.93±31.74 16.38±6.95 13.81±6.19 11.89±9.59 11.55±9.18 16.51±7.46
L-DM 104.35±70.92 99.60±72.18 50.64±39.06 19.51±11.81 29.48±21.89 19.89±15.59 21.21±17.34 23.07±12.63
L-DH 58.37±45.66 50.90±36.05 28.56±13.00 33.23±16.66 50.73±32.19 37.22±29.52 37.95±27.33 37.95±20.04
F 9.789 14.349 15.369 13.962 16.323 11.052 12.514 11.722
P <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001
组别 CNR值
右肾盂 左肾盂 右输尿管 左输尿管 膀胱
S-AV60 89.85±39.59 95.49±49.70 85.39±37.27 87.29±40.15 61.14±33.82
L-AV60 98.23±39.32 105.06±40.87 95.79±33.73 95.95±35.25 59.23±27.38
L-DM 110.24±42.90 119.43±45.29 107.21±37.18 109.19±40.59 68.18±32.82
L-DH 149.36±62.83 160.10±62.62 144.21±51.17 145.43±54.25 91.34±42.29
F 6.755 9.620 12.085 10.571 5.492
P <0.001 <0.001 <0.001 <0.001 0.001
图1 不同组别CTU的最大强度投影(MIP)重建图注:4组图为MIP重建,从左到右依次为S组60%ASIR-V、L组60%ASIR-V、DLIR-M、DLIR-H。S组60%ASIR-V、L组60%ASIR-V噪声点密集,DLIR-H,DLIR-M噪声最佳。S组60%ASIR-V图像在输尿管、膀胱边缘模糊,而DLIR-M、DLIR-H边缘清晰锐利,无颗粒感;MIP为最大强度投影,VR为容积再现,DLIR-M为中等级别深度学习图像重建,DLIR-H为高等级别深度学习图像重建,ASIR-V为自适应统计迭代重建-V
表3 不同组别CTU重建图像主观评分比较
图2 CTU重建图像不同组别的硬化伪影注:3组图分别为L组同一患者同层轴位图像,其中L-DH的硬化伪影面积明显小于L-AV60、L-DM。两名医师对该图像在硬化伪影的主观评分分别为(3,3,4) (3,4,5)
[1]
Li X, Shu H, Zhang Y, et al. Low-dose CT with adaptive statistical iterative reconstruction for evaluation of urinary stone[J]. Oncotarget, 2018, 9(28): 20103-20111.
[2]
Ren Z, Zhang X, Hu Z, et al. Reducing radiation dose and improving image quality in ct portal venography using 80 kv and adaptive statistical iterative reconstruction-v in slender patients[J]. Acad Radiol, 2020, 27(2): 233-243.
[3]
Lee HN, Lee SM, Choe J, et al. Diagnostic performance of CT-guided percutaneous transthoracic core needle biopsy using low tube voltage (100 kVp): comparison with conventional tube voltage (120 kVp). Acta Radiol[J]. 2018, 59(4): 425-433.
[4]
Goodenberger MH, Wagner-Bartak NA, Gupta S, et al. Computed tomography image quality evaluation of a new iterative reconstruction algorithm in the abdomen (adaptive statistical iterative reconstruction-v) a comparison with model-based iterative reconstruction, adaptive statistical iterative reconstruction, and filtered back projection reconstructions[J]. J Comput Assist Tomogr, 2018, 42(2): 184-190.
[5]
Park C, Choo KS, Kim JH, et al. Image quality and radiation dose in ct venography using model-based iterative reconstruction at 80 kvp versus adaptive statistical iterative reconstruction-Vat 70 kVp[J]. Korean J Radiol, 2019, 20(7): 1167-1175.
[6]
Goodenberger MH, Wagner-Bartak NA, Gupta S, et al. Computed tomography image quality evaluation of a new iterative reconstruction algorithm in the abdomen (adaptive statistical iterative reconstruction-v) a comparison with model-based iterative reconstruction, adaptive statistical iterative reconstruction, and filtered back projection reconstructions[J]. J Comput Assist Tomogr, 2018, 42(2): 184-190.
[7]
Laurent G, Villani N, Hossu G, et al. Full model-based iterative reconstruction (MBIR) in abdominal CT increases objective image quality, but decreases subjective acceptance[J]. Eur Radiol, 2019, 29(8): 4016-4025.
[8]
Jensen CT, Liu X, Tamm EP, et al. Image quality assessment of abdominal ct by use of new deep learning image reconstruction: initial experience[J]. AJR Am J Roentgenol, 2020, 215(1): 50-57.
[9]
Hwang I, Cho JY, Kim SY, et al. Low tube voltage computed tomography urography using low-concentration contrast media: Comparison of image quality in conventional computed tomography urography[J]. Eur J Radiol, 2015, 84(12): 2454-2463.
[10]
Kim SY, Cho JY, Lee J, et al. Low-Tube-Voltage CT urography using low-concentration-iodine contrast media and iterative reconstruction: a multi-institutional randomized controlled trial for comparison with conventional CT urography[J]. Korean J Radiol, 2018, 19(6): 1119-1129.
[11]
Kwon H, Cho J, Oh J, et al. The adaptive statistical iterative reconstruction-V technique for radiation dose reduction in abdominal CT: comparison with the adaptive statistical iterative reconstruction technique[J]. Br J Radiol, 2015, 88(1054): 20150463.
[12]
Arndt C, Güttler F, Heinrich A, et al. Deep learning ct image reconstruction in clinical practice[J]. Rofo, 2021, 193(3): 252-261.
[13]
Willemink MJ, Noël PB. The evolution of image reconstruction for CT-from filtered back projection to artificial intelligence[J]. Eur Radiol, 2019, 29(5): 2185-2195.
[14]
Kim JH, Yoon HJ, Lee E, et al. Validation of deep-learning image reconstruction for low-dose chest computed tomography scan: emphasis on image quality and noise[J]. Korean J Radiol, 2021, 22(1): 131-138.
[15]
王甜, 曹治婷, 韩秋丽. 低管电压联合低剂量对比剂在肺动脉CT成像中的应用[J]. 实用医学杂志, 2016, 32(15): 2532-2534.
[1] 张梅芳, 谭莹, 朱巧珍, 温昕, 袁鹰, 秦越, 郭洪波, 侯伶秀, 黄文兰, 彭桂艳, 李胜利. 早孕期胎儿头臀长正中矢状切面超声图像的人工智能质控研究[J]. 中华医学超声杂志(电子版), 2023, 20(09): 945-950.
[2] 杨菲菲, 林锡祥, 陈亦新, 王秋霜, 张丽伟, 陈煦, 张梅青, 王淑华, 何昆仑. 基于深度学习的超声心动图自动识别节段性室壁运动异常的研究[J]. 中华医学超声杂志(电子版), 2023, 20(04): 424-429.
[3] 唐玮, 何融泉, 黄素宁. 深度学习在乳腺癌影像诊疗和预后预测中的应用[J]. 中华乳腺病杂志(电子版), 2023, 17(06): 323-328.
[4] 李锐颖, 危望, 王达志, 时志斌. 深度学习技术在膝关节疾病中的研究现状与展望[J]. 中华关节外科杂志(电子版), 2023, 17(05): 722-725.
[5] 范帅华, 郭伟, 郭军. 基于机器学习的决策树算法在血流感染预后预测中应用现状及展望[J]. 中华实验和临床感染病杂志(电子版), 2023, 17(05): 289-293.
[6] 张辉, 蔡敏, 黄湘雅. 数字化技术和人工智能在上颌窦底提升术的临床应用[J]. 中华口腔医学研究杂志(电子版), 2023, 17(04): 244-252.
[7] 李晓阳, 刘柏隆, 周祥福. 大数据及人工智能对女性盆底功能障碍性疾病的诊断及风险预测[J]. 中华腔镜泌尿外科杂志(电子版), 2023, 17(06): 549-552.
[8] 胡博文, 戴英波. 泌尿外科机器人手术新趋势[J]. 中华腔镜泌尿外科杂志(电子版), 2023, 17(04): 418-421.
[9] 李腾成, 狄金明. 2023 V1版前列腺癌NCCN指南更新要点解读[J]. 中华腔镜泌尿外科杂志(电子版), 2023, 17(04): 313-318.
[10] 邢晓伟, 刘雨辰, 赵冰, 王明刚. 基于术前腹部CT的卷积神经网络对腹壁切口疝术后复发预测价值[J]. 中华疝和腹壁外科杂志(电子版), 2023, 17(06): 677-681.
[11] 雷漫诗, 邓锶锶, 汪昕蓉, 黄锦彬, 向青, 熊安妮, 孟占鳌. 人工智能辅助压缩感知技术在上腹部T2WI压脂序列中的应用[J]. 中华肝脏外科手术学电子杂志, 2023, 12(05): 551-556.
[12] 韩冰, 顾劲扬. 深度学习神经网络在肝癌诊疗中的研究及应用前景[J]. 中华肝脏外科手术学电子杂志, 2023, 12(05): 480-485.
[13] 葛云鹏, 崔红元, 宋京海. 人工智能在原发性肝癌诊断、治疗及预后中的应用[J]. 中华肝脏外科手术学电子杂志, 2023, 12(04): 367-371.
[14] 王晓东, 汪恺, 葛昭, 丁忠祥, 徐骁. 计算机视觉技术在肝癌肝移植疗效提升中的研究进展[J]. 中华肝脏外科手术学电子杂志, 2023, 12(04): 361-366.
[15] 李京珂, 张妍春, 武佳懿, 任秀瑜. 深度学习在糖尿病视网膜病变筛查、评级及管理中的研究进展[J]. 中华眼科医学杂志(电子版), 2023, 13(04): 241-246.
阅读次数
全文


摘要