Application of Deep Learning Reconstruction Algorithm in Upper Abdomen CT
-
摘要: 目的：通过分析比较自适应统计迭代重建（ASIR）算法和深度学习重建（DLIR）算法在上腹部CT成像中的图像质量，探讨DLIR算法在上腹部CT成像中的应用价值。方法：回顾性纳入75例患者上腹部CT平扫图像，利用自适应统计迭代重建算法ASIR（30％、50％、70％、90％）和深度学习重建算法（DL-L、DL-M、DL-H）重建图像，共7组。测量每组图像肝脏、胰腺、竖脊肌的CT值和SD值，并计算信噪比（SNR）和对比噪声比（CNR），采用单因素方差分析对各指标进行客观评价；同时由两位放射医师对图像质量和噪声评分，采用Friedman <i<M</i<检验进行比较。结果：①七组重建图像的SD值、SNR、肝脏CNR差异有统计学意义。②DL-L与ASIR 50％、DL-M与ASIR 70％、DL-H与ASIR 90％ 间各ROI处CT值、SD值、SNR值、CNR值无差异。③三种深度学习重建算法间随等级升高，SNR值升高，差异有统计学意义；且DL-H 算法的SNR值高于ASIR 30％、ASIR 50％，SD值低于除ASIR 90％ 外的其余5组重建算法。④七组图像主观评分差异有统计学意义，算法DL-H具有最佳的图像质量和最低的噪声，DL-M、ASIR 90％、DL-L、ASIR 70％、ASIR 50％、ASIR 30％ 图像噪声依次增加。结论：深度学习重建算法能够降低上腹部图像噪声，提高图像质量，且随等级升高，图像噪声降低、质量提高、信噪比升高。Abstract: Objective: To explore the application of deep learning image reconstruction (DLIR) algorithm in upper abdominal CT imaging by analyzing the image quality of adaptive statistical iterative reconstruction (ASIR) algorithm and DLIR. Methods: Retrospectively included 75 patients’ upper abdominal CT plain scan images, using adaptive statistical iterative reconstruction algorithm ASIR (30%, 50%, 70%, 90%) and deep learning reconstruction algorithm (DL-L, DL-M, DL-H) to reconstruct images, a total of 7 groups. Measured the CT and SD values of the liver, pancreas, and erector spinae , and calculated the signal to noise ratio (SNR) and contrast to noise ratio (CNR). Objective indicators were evaluated by one-way ANOVA. Two radiologists scored the image quality and noise, and compared them with Friedman <i<M</i< test. Results: (1) The SD value, SNR, and liver CNR of the seven reconstructed images had statistically significant differences. (2) The difference in CT value, SD value, SNR value and CNR value at each ROI between DL-L and ASIR 50%, DL-M and ASIR 70%, DL-H and ASIR 90% was small. (3) The SNR value of the three DLIR algorithms increased as the level increased, and the difference was statistically significant; and the SNR value of the DL-H algorithm was higher than ASIR 30% and ASIR 50%, and the SD value was lower than the other five reconstruction algorithms except for the ASIR 90%. (4) The difference in the subjective scores of the seven groups of images was statistically significant. The algorithm DL-H had the best image quality and the lowest noise, DL-M, ASIR 90%, DL-L, ASIR 70%, ASIR 50%, ASIR 30% image noise in sequence increased. Conclusion: The DLIR algorithm can reduce the image noise of the upper abdomen and improve the image quality. As the level increased, the image noise decreased, the quality improved, and the signal-to-noise ratio increased.
-
-
![]()
图 1 上腹部7组重建图像
（a）～（g）依次是算法DL-L、DL-M、DL-H、ASIR 30％、ASIR 50％、ASIR 70％、ASIR 90％。
Figure 1. 7 groups of reconstructed images of upper abdomen
![]()
图 2 上腹部7组重建图像局部放大图
（a）～（g）依次是算法DL-L、DL-M、DL-H、ASIR 30％、ASIR 50％、ASIR 70％、ASIR 90％。
Figure 2. Partial enlarged view of 7 groups of reconstructed images of upper abdomen
表 1 图像主观评价5级评分标准
Table 1 5-level scoring standard for subjective image evaluation
评分 评分细节 5分 肝脏、胰腺等上腹部组织结构显示非常清晰，图像细腻，能提供充分的诊断信息 4分 肝脏、胰腺等上腹部组织结构显示较为清晰，图像较细腻，能提供足够的诊断信息 3分 肝脏、胰腺等上腹部组织结构显示欠清晰，图像欠细腻，能提供一定的诊断信息 2分 肝脏、胰腺等上腹部组织结构显示模糊，图像粗糙，图像提供的诊断信息不足 1分 肝脏、胰腺等上腹部组织结构无法清晰显示，伪影重，图像不清，不能提供诊断信息 
下载: 导出CSV
表 2 七组重建图像的CT值、SD值分析
Table 2 Analysis of CT value and SD value of seven groups of reconstructed images
算法 CT值/HU SD值/HU ROI1 ROI2 ROI3 ROI1 ROI2 ROI3 DL-L 60.65±8.64 44.63±7.45 49.43±8.70 14.05±4.21 15.38±4.73 15.25±5.57 DL-M 61.24±9.62 44.33±7.46 49.46±8.67 11.38±3.27 12.81±3.94 12.68±4.97 DL-H 60.64±8.43 44.33±7.27 49.48±8.66 8.41±2.15 9.53±3.16 9.90±4.54 ASIR 30％ 60.58±8.80 44.50±7.68 49.46±8.55 17.70±5.25 19.34±5.54 18.79±6.26 ASIR 50％ 60.65±8.81 44.29±7.48 49.45±8.60 14.69±4.33 16.14±4.95 15.77±5.46 ASIR 70％ 60.59±8.96 44.32±7.28 49.51±8.66 11.86±4.43 13.07±4.24 12.85±4.92 ASIR 90％ 60.75±8.70 44.29±7.53 49.55±8.88 8.74±2.83 10.1±3.57 10.27±4.70 P >0.05 >0.05 >0.05 <0.05 <0.05 <0.05
下载: 导出CSV
表 3 七组重建图像的SNR值、CNR值分析
Table 3 Analysis of SNR and CNR values of seven groups of reconstructed images
算法 SNR CNR ROI1 ROI2 ROI3 ROI1 ROI2 DL-L 4.67±1.53 3.18±1.05 3.67±1.30 0.71±0.70 -0.41±0.52 DL-M 5.80±1.91 3.80±1.28 4.47±1.64 0.90±0.89 -0.53±0.64 DL-H 7.66±2.30 5.11±1.64 5.85±2.17 1.07±1.03 -0.70±0.80 ASIR 30％ 3.71±1.28 2.51±0.86 2.94±1.03 0.58±0.58 -0.33±0.42 ASIR 50％ 4.49±1.58 3.02±1.04 3.54±1.29 0.70±0.67 -0.42±0.51 ASIR 70％ 5.67±2.09 4.47±1.76 4.42±1.67 0.84±0.86 -0.22±0.63 ASIR 90％ 7.70±2.94 4.96±1.83 5.79±2.48 1.09±1.05 -0.72±0.84 P <0.05 <0.05 <0.05 <0.05 >0.05
下载: 导出CSV
表 4 图像质量主观评分分析
Table 4 Image quality subjective scoring analysis
评分内容 评分者 DL-L DL-M DL-H ASIR 30％ ASIR 50％ ASIR 70％ ASIR 90％ 质量 Reader 1 4.41±0.59 4.79±0.41 4.97±0.16 3.57±0.68 3.89±0.68 4.42±0.57 4.71±0.45 Reader 2 4.30±0.56 4.84±0.36 4.96±0.19 3.53±0.57 3.87±0.61 4.36±0.48 4.75±0.43 噪声 Reader 1 2.21±0.57 1.18±0.39 1.01±0.11 3.79±0.41 3.37±0.56 2.66±0.55 1.83±0.50 Reader 2 2.18±0.51 1.11±0.31 1.05±0.22 3.91±0.29 3.18±0.39 2.36±0.55 1.68±0.46
下载: 导出CSV
-
[1] 张卓璐, 王征, 刘卓, 等. 迭代重建算法对冠状动脉Agatston钙化积分的影响[J]. 临床放射学杂志, 2020,39(10): 2093−2097. ZHANG Z L, WANG Z, LIU Z, et al. Influence of iterative reconstruction algorithm on coronary artery agatston calcium score[J]. Journal of Clinical Radiology, 2020, 39(10): 2093−2097. (in Chinese).
[2] 张喜荣, 贺太平, 贾永军, 等. 低辐射剂量下FBP、ASIR和ASIR-V 3种不同重建算法对上腹部CT图像质量的影响[J]. 中国中西医结合影像学杂志, 2020,18(3): 305−308. doi: 10.3969/j.issn.1672-0512.2020.03.027 ZHANG X R, HE T P, JIA Y J, et al. Effect of three different reconstruction algorithms (FBP, ASIR and ASIR-V) on the upper abdominal CT image quality with low radiation dose[J]. Chinese Imaging Journal of Integrated Traditional and Western Medicine, 2020, 18(3): 305−308. (in Chinese). doi: 10.3969/j.issn.1672-0512.2020.03.027
[3] 贾永军, 于勇, 贺太平, 等. 新一代基于模型的迭代重建在低剂量上腹部CT中的应用[J]. 中国医学影像技术, 2017,33(12): 1882−1887. JIA Y J, YU Y, HE T P, et al. Application of new model-based iterative reconstruction in low-dose upper abdominal CT[J]. Chinese Journal of Medical Imaging Technology, 2017, 33(12): 1882−1887. (in Chinese).
[4] 陈其锋, 林梓朗, 杨宇凌. 低剂量扫描联合迭代重建CTA在颈部疾病患者中的应用效果及价值研究[J]. 医学理论与实践, 2021,34(7): 1205−1207. [5] 姜一, 秦立新, 李宝学, 等. 80kV结合低剂量对比剂和迭代重建在胸部CT增强检查中的运用[J]. 中国医疗设备, 2021,36(3): 102−105. doi: 10.3969/j.issn.1674-1633.2021.03.022 JIANG Y, QIN L X, LI B X, et al. Application of 80kV Combined with low-dose contrast medium and iterative reconstruction in enhanced chest CT examination[J]. China Medical Devices, 2021, 36(3): 102−105. (in Chinese). doi: 10.3969/j.issn.1674-1633.2021.03.022
[6] FRANCK C, ZHANG G, DEAK P, et al. Preserving image texture while reducing radiation dose with a deep learning image reconstruction algorithm in chest CT: A phantom study[J]. Physica Medica, 2021, 81: 86−93. doi: 10.1016/j.ejmp.2020.12.005
[7] BENZ D C, BENETOS G, RAMPIDIS G, et al. Validation of deep-learning image reconstruction for coronary computed tomography angiography: Impact on noise, image quality and diagnostic accuracy[J]. Journal of Cardiovascular Computed Tomography, 2020, 14(5): 444−451. doi: 10.1016/j.jcct.2020.01.002
[8] PARAKH A, CAO J, PIERCE T T, et al. Sinogram-based deep learning image reconstruction technique in abdominal CT: Image quality considerations[J]. European Radiology, 2021: 1−12.
[9] NJØLSTAD T, SCHULZ A, GODT J C, et al. Improved image quality in abdominal computed tomography reconstructed with a novel deep learning image reconstruction technique: Initial clinical experience[J]. Acta Radiologica Open, 2021, 10(4): 20584601211008391.
[10] 贾秀川, 陈英敏, 暴云锋, 等. 双源CT小肠造影双能量虚拟平扫与常规平扫对比研究[J]. 中国医疗设备, 2021,36(1): 87−89, 93. doi: 10.3969/j.issn.1674-1633.2021.01.018 JIA X C, CHEN Y M, BAO Y F, et al. Comparative study of virtual and conventional non-contrast CT enterography using dual-source and energy CT[J]. China Medical Devices, 2021, 36(1): 87−89, 93. (in Chinese). doi: 10.3969/j.issn.1674-1633.2021.01.018
[11] 曾文, 曾令明, 徐旭, 等. 基于深度学习的图像重建算法在胸部薄层CT中的降噪效果评估[J]. 四川大学学报(医学版), 2021,52(2): 286−292. ZENG W, ZENG L M, XU X, et al. Noise reduction effect of deep-learning-based image reconstruction algorithms in thin-section chest CT[J]. Journal of Sichuan University (Medical Sciences), 2021, 52(2): 286−292. (in Chinese).
[12] SILVA A C, LAWDER H J, HARA A, et al. Innovations in CT dose reduction strategy: Application of the adaptive statistical iterative reconstruction algorithm[J]. American Journal of Roentgenology, 2010, 194(1): 191−199. doi: 10.2214/AJR.09.2953
[13] JENSEN C T, WAGNER-BARTAK N A, VU L N, et al. Detection of colorectal hepatic metastases is superior at standard radiation dose CT versus reduced dose CT[J]. Radiology, 2019, 290(2): 400−409. doi: 10.1148/radiol.2018181657
[14] LI L L, WANG H, SONG J, et al. A feasibility study of realizing low-dose abdominal CT using deep learning image reconstruction algorithm[J]. Journal of X-ray Science and Technology, 2021, 29(2): 361−372. doi: 10.3233/XST-200826
[15] 孙记航, 王帆宁, 段晓岷, 等. 自适应迭代重建技术结合高分辨算法提高儿童低剂量胸部CT肺脏病变显示的能力[J]. 中国医学影像技术, 2017,33(5): 773−777. SUN J H, WANG F N, DUAN X M, et al. Improve image resolution in low-dose pediatric chest CT scans with combination of adaptive statistical iterative reconstruction and sharp recon kernel[J]. Chinese Journal of Medical Imaging Technology, 2017, 33(5): 773−777. (in Chinese).
[16] HATA A, YANAGAWA M, YOSHIDA Y, et al. The image quality of deep-learning image reconstruction of chest CT images on a mediastinal window setting[J]. Clinical Radiology, 2021, 76(2): 155.e15−155.e23. doi: 10.1016/j.crad.2020.10.011
[17] SINGH R, DIGUMARTHY S R, MUSE V V, et al. Image quality and lesion detection on deep learning reconstruction and iterative reconstruction of submillisievert chest and abdominal CT[J]. American Journal of Roentgenology, 2020, 214(3): 566−573. doi: 10.2214/AJR.19.21809
[18] HIGAKI T, NAKAMURA Y, ZHOU J, et al. Deep learning reconstruction at CT: Phantom study of the image characteristics[J]. Academic radiology, 2020, 27(1): 82−87. doi: 10.1016/j.acra.2019.09.008
[19] KIM J H, YOON H J, 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 Journal of Radiology, 2021, 22(1): 131. doi: 10.3348/kjr.2020.0116
[20] NODA Y, KAGA T, KAWAI N, et al. Low-dose whole-body CT using deep learning image reconstruction: Image quality and lesion detection[J]. The British Journal of Radiology, 2021, 94: 20201329. doi: 10.1259/bjr.20201329
[21] NAM J G, HONG J H, KIM D S, et al. Deep learning reconstruction for contrast-enhanced CT of the upper abdomen: Similar image quality with lower radiation dose in direct comparison with iterative reconstruction[J]. European Radiology, 2021, 31(8): 5533−5543. doi: 10.1007/s00330-021-07712-4
[22] CHENG Y, HAN Y, LI J, et al. Low-dose CT urography using deep learning image reconstruction: A prospective study for comparison with conventional CT urography[J]. The British Journal of Radiology, 2021, 94(1120): 20201291. doi: 10.1259/bjr.20201291
计量
- 文章访问数: 472
- HTML全文浏览量: 322
- PDF下载量: 78
