Impact of noise index combined with deep learning image reconstruction on image quality and radiation dose
-
摘要:
目的 探讨深度学习图像重建(DLIR)在超低剂量肺部CT成像中的应用价值。 方法 选取2024年3~4月在陕西中医药大学附属医院行肺部CT平扫患者66例。所有患者均采用GE Revolution CT扫描，固定管电压100 kVp，第1次采用噪声指数(NI)=15的常规辐射剂量扫描，滤波反投影算法重建图像；第2次采用NI=45的超低辐射剂量扫描，中、高等强度深度学习图像重建(DLIR-M、DLIR-H)进行对比。在3组重建图像上测量左上肺乏血供区域CT值与标准差值(SD)，SD代表噪声，计算信噪比(SNR)。由2位放射科诊断医师采用5分法进行主观评价，比较3组客观数值和主观评分。 结果 NI=45组约减少93.7%辐射剂量；DLIR强度影响超低剂量条件下客观指标，DLIR-H较DLIR-M有更低的噪声，更高的SNR(P < 0.05)；2位医师对3组图像质量一致性评价好(Kappa值为0.952、0.846、0.903)；对比3组图像质量评分、图像合格率及满意率，差异无统计学意义(P > 0.05)。 结论 在减少93.7%辐射剂量条件下，DLIR能够获得与常规剂量接近的肺部图像，进一步减低了肺部疾病筛查的辐射剂量。 Abstract:Objective To explore the application value of Deep learning image reconstruction (DLIR) in ultra-low-dose chest CT imaging. Methods A total of 66 patients with chest CT scans in the Affiliated Hospital of Shaanxi University of Traditional Chinese Medicine from March to April 2024 were collected.All patients were used GE Revolution CT scans, the fixed tube voltage was 100 kVp, and the first with conventional radiation dose with noise index (NI) =15, and filtered back projection reconstructed images; the second was scanned with ultra-low-dose with NI= 45, and medium and high intensity deep learning image reconstruction (DLIR- M、DLIR- H) were compared. The CT value and standard deviation (SD) of the left upper pulmonary hypovascular region were measured on three reconstructed images, SD represented noise, and the signal-to-noise ratio (SNR) was calculated. Subjective evaluation of 5-point method was used by two radiologists. The objective value and subjective score of three reconstructed images were compared. Results The NI=45 group reducted the radiation dose by 93.7%. The intensity of DLIR affected the objective value under ultra-low-dose condition, DLIR-H resulted in ower noise and higher SNR than DLIR-M (P < 0.05). Two physicians evaluated the image quality consistency of the three reconstructed images (Kappa= 0.952, 0.846, 0.903). The image quality scores, pass rates and satisfaction rates had no significant differences between three groups (P > 0.05). Conclusion Under the condition of reducing the radiation dose by 93.7%, DLIR can obtain images of the lung that are close to the conventional radiation dose, and the radiation dose for lung disease screening has been further reduced. -
Key words:
- ultra-low dose /
- lung CT /
- deep learning /
- image quality /
- radiation dose
-
图 1 3组重建图像
Figure 1. Three groups of reconstructed images. A: FBP; B: DLIR-M+E3; C: DLIR-H+E3.
表 1 3组图像CT值、SD值及SNR值比较
Table 1. Comparison of CT, SD and SNR values of three groups of images
Item FBP DLIR-M+E3 DLIR-H+E3 F/H P CT -895.98±17.41 -892.83±18.36 -892.47±18.20 0.761 0.469 SD 67.80±5.62 73.12±6.18 63.89±5.17 44.005 < 0.001 SNR 13.43(1.67) 12.12(1.53) 14.06(1.87) 55.084 < 0.001 FBP: Filtered back projection; DLIR-M + E3/DLIR-H + E3: Deep learning image reconstruction at medium/high strength with additional post-processing using an edge-enhancement filter E3; SD: Standard deviation; SNR: Signal to noise ratio. 
下载: 导出CSV
表 2 3组图像质量主观评分一致性分析
Table 2. Consistency analysis of subjective scores of image quality in three groups (n)
Image groups Evaluator Image quality score 1 2 3 4 5 FBP A 0 4 34 20 8 B 0 5 34 19 8 DLIR-M+E3 A 0 7 38 17 4 B 0 8 38 16 4 DLIR-H+E3 A 0 7 34 19 6 B 0 6 36 18 6
下载: 导出CSV
表 3 3组图像质量主观评分比较
Table 3. Comparison of subjective image quality scores among the three groups (n)
Image groups Image quality score < 3 3-4 ≥4 FBP 5 34 27 DLIR-M+E3 8 38 20 DLIR-H+E3 5 37 24 H 1.97 P 0.373
下载: 导出CSV
表 4 3组图像合格率、满意率比较
Table 4. Comparison of image pass rate and satisfaction rate among the three groups (%)
Image groups Passing rate Satisfaction rate FBP 92.4(61/66) 40.9(27/66) DLIR-M+E3 87.9(58/66) 30.3(20/66) DLIR-H+E3 90.9(60/66) 36.4(24/66) χ2 0.815 1.625 P 0.665 0.444
下载: 导出CSV
-
[1] 王泽润, 曹永佩, 杨东升, 等. 管电压70 kV下迭代重建技术对超低剂量胸部CT扫描图像质量的影响[J]. 宁夏医学杂志, 2022, 44(03): 249-51. [2] Hauptmann M, Daniels RD, Cardis E, et al. Epidemiological studies of low-dose ionizing radiation and cancer: summary bias assessment and meta-analysis[J]. J Natl Cancer Inst Monogr, 2020, 2020(56): 188-200. doi: 10.1093/jncimonographs/lgaa010 [3] Friedman-Jimenez G, Kato I, Factor-Litvak P, et al. Low-dose ionizing radiation and cancer mortality among enlisted men stationed on nuclear-powered submarines in the United States Navy [J]. Int J Radiat Biol, 2022, 98(10): 1542-50. doi: 10.1080/09553002.2022.2055805 [4] Wang XC, Tian LL, Fan CX, et al. The adaptive responses in nonsmall cell lung cancer A549 cell lines induced by low-dose ionizing radiation and the variations of miRNA expression[J]. Dose Response, 2021, 19(4): 15593258211039931. doi: 10.1177/15593258211039931 [5] Lowenstein LM, Deyter GMR, Nishi S, et al. Shared decisionmaking conversations and smoking cessation interventions: critical components of low-dose CT lung cancer screening programs[J]. Transl Lung Cancer Res, 2018, 7(3): 254-71. doi: 10.21037/tlcr.2018.05.10 [6] Mileto A, Guimaraes LS, McCollough CH, et al. State of the art in abdominal CT: the limits of iterative reconstruction algorithms[J]. Radiology, 2019, 293(3): 491-503. doi: 10.1148/radiol.2019191422 [7] 李杨, 王峰先, 李洋, 等. TrueFidelity深度学习图像重组技术在儿童副鼻窦超低剂量CT检查中的初步应用研究[J]. 临床放射学杂志, 2023, 42(6): 1003-8. [8] 金鑫, 陈杭美, 傅钰沁. 深度学习图像重建算法在低剂量扫描对肺结节显示及测量的影响[J]. 医学影像学杂志, 2024, 34(3): 140-2. [9] 郭晓强, 侯楠楠, 王晓龙. 深度学习重建对胸部低剂量CT影像的影响研究[J]. 中国卫生工程学, 2024, 23(2): 232-4. [10] 刘海法, 杨丽, 王琦, 等. 深度学习重建算法对胸部低剂量CT肺结节测量及显示影响的模体研究[J]. 放射学实践, 2023, 38(8): 977-84. [11] 宋兰, 田杜雪, 王金华, 等. 胸部超低剂量CT应用深度学习重建行肺癌筛查的可行性研究[J]. 中华放射学杂志, 2022, 56(6): 667-72. doi: 10.3760/cma.j.cn112149-20210810-00463 [12] 衣闯, 梁妍, 陈志仁. 低剂量CT胸部体检中噪声指数的设定[J]. 中国医学影像学杂志, 2017, 25(10): 756-8, 760. doi: 10.3969/j.issn.1005-5185.2017.10.010 [13] 刘铁, 李艳艳, 李鹏, 等. 胸部CT图像质量评价与辐射剂量的临床应用研究[J]. 影像技术, 2024, 36(1): 19-23, 38. doi: 10.3969/j.issn.1001-0270.2024.01.04 [14] 凌问非, 瞿亚伟, 余兆祥, 等. 肺部低剂量CT扫描对图像质量和辐射剂量的影响[J]. 影像研究与医学应用, 2023, 7(13): 96-8. doi: 10.3969/j.issn.2096-3807.2023.13.029 [15] Arns A, Wertz H, Boda-Heggemann J, et al. Ultrafast single breathhold cone-beam CT lung cancer imaging with faster linac gantry rotation[J]. Radiother Oncol, 2019, 135: 78-85. doi: 10.1016/j.radonc.2019.02.004 [16] Gierada DS, Black WC, Chiles C, et al. Low-dose CT screening for lung cancer: evidence from 2 decades of study[J]. Radiol Imaging Cancer, 2020, 2(2): e190058. doi: 10.1148/rycan.2020190058 [17] 任心爽, 安云强, 吕滨, 等. 中国冠状动脉CT血管成像扫描技术及辐射剂量的现状调查[J]. 中华放射学杂志, 2022, 56(4): 405-10. doi: 10.3760/cma.j.cn112149-20210720-00689 [18] 李中丹, 曾绪知, 董科, 等. 低剂量胸部CT检查对胸部肿瘤性病变的检出和诊断效果评价[J]. 医学影像学杂志, 2024, 34(5): 66-9. [19] 陈伟婷, 马光明, 何立宇, 等. 超低剂量胸部CT不同重建算法对肺炎定量及图像质量的影响[J]. 分子影像学杂志, 2024, 47(3): 282-8. doi: 10.12122/j.issn.1674-4500.2024.03.10 [20] 吕东博. 超低剂量下基于重建算法对胸部CT图像质量及肺结节的仿真体模研究[D]. 郑州: 郑州大学, 2021. [21] 冯倩倩, 于卫军, 唐威, 等. 三种不同重建算法在低剂量胸部CT图像质量中的比较[J]. 中国肿瘤临床与康复, 2021, 28(12): 1409-13. [22] Brady SL, Trout AT, Somasundaram E, et al. Improving image quality and reducing radiation dose for pediatric CT by using deep learning reconstruction[J]. Radiology, 2021, 298(1): 180-8. doi: 10.1148/radiol.2020202317 [23] 王梅, 花立春, 唐颖, 等. 基于深度学习的图像重建算法在CT图像重建中的应用进展[J]. 中国医疗设备, 2022, 37(12): 157-61. doi: 10.3969/j.issn.1674-1633.2022.12.032 [24] Tschauner S, Zellner M, Pistorius S, et al. Ultra-low-dose lung multidetector computed tomography in children-Approaching 0.2 millisievert[J]. Eur J Radiol, 2021, 139: 109699. doi: 10.1016/j.ejrad.2021.109699 [25] Wang YN, Du Y, Shi GF, et al. A preliminary evaluation study of applying a deep learning image reconstruction algorithm in lowkilovolt scanning of upper abdomen[J]. J Xray Sci Technol, 2021, 29(4): 687-95. [26] 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): 286-92. [27] Svalkvist A, Fagman E, Vikgren J, et al. Evaluation of deeplearning image reconstruction for chest CT examinations at two different dose levels[J]. J Appl Clin Med Phys, 2023, 24(3): e13871. doi: 10.1002/acm2.13871 [28] Jiang BB, Li NY, Shi XM, et al. Deep learning reconstruction shows better lung nodule detection for ultra-low-dose chest CT[J]. Radiology, 2022, 303(1): 202-12. doi: 10.1148/radiol.210551 [29] 田骞, 张喜荣, 贾永军, 等. ASIR结合自动调节管电流技术在降低胸部CT辐射剂量的研究[J]. 现代医用影像学, 2021, 30(10): 1820-5. doi: 10.3969/j.issn.1006-7035.2021.10.008 [30] 张卓璐, 陈菁, 刘卓, 等. 深度学习重建算法对肺部CT定量分析及图像质量的影响[J]. 放射学实践, 2023, 38(4): 434-40. -
点击查看大图
图(1) / 表(4)
计量
- 文章访问数: 51
- HTML全文浏览量: 15
- PDF下载量: 2
- 被引次数: 0
