摘要
目的探讨深度学习重建算法对模拟肺结节定量分析准确性及图像质量的影响。方法使用GE Revolution CT对内置9枚模拟结节的仿真胸部体模进行低剂量CT扫描。采用滤波反投影法(Filtered Back Projection,FBP)及不同混合权重的深度学习图像重建(Deep Learning Image Reconstruction,DLIR)和自适应统计迭代重建(Adaptive Statistical Iterative Reconstruction-Veo,ASiR-V)算法对图像进行重建。利用人工智能辅助诊断软件对所有图像进行自动测量,与手动测量结果进行对比,并计算信噪比(Signal to Noise Ratio,SNR)及对比噪声比(Contrast to Noise Ratio,CNR)。结果不同混合权重、不同算法均不影响肺结节的手动测量结果(P>0.05),但DLIR的使用可能影响肺结节大小的自动测量结果(P=0.01);此外,DLIR算法计算的SNR、CNR最高,且DLIR-H拥有最好的图像质量(P<0.05)。结论与ASiR-V和FBP算法相比,DLIR可以显著改善图像质量;且DLIR的混合权重越高,图像质量改善越明显,测量方式不同不会影响这一结果。
Objective To explore the influence of deep learning reconstruction algorithm on quantitative analysis accuracy and image quality of simulated pulmonary nodules.Methods Low-dose CT scanning of a simulated chest model with nine simulated nodules was performed using GE Revolution CT.Filtered back projection(FBP),and different mixed weights deep learning image reconstruction(DLIR)algorithms and adaptive statistical iterative reconstruction-Veo(ASIR-V)algorithms were used for image reconstruction.All images were automatically measured by an AI-assisted diagnostic software,and the results were compared with manual measurements.The signal to noise ratio(SNR)and contrast to noise ratio(CNR)were calculated.Results Different weights and algorithm both had no effect on the manual measurement of nodule diameter(P>0.05);but DLIR algorithm might result in a statistically significant deviation between the automatic measurement of nodules and the true value(P=0.01).The SNR and CNR calculated by the DLIR algorithm were the highest,and DLIR-H had the best image quality(P<0.05).Conclusion Compared with ASiR-V and FBP algorithms,DLIR algorithm could improve image quality significantly.The higher the mixed weight of DLIR,the more obvious the improvement of image quality,and the different measurement methods will not affect this result.
作者
邓蕾
郭宝斌
姚悦
杨全新
李晓会
DENG Lei;GUO Baobin;YAO Yue;YANG Quanxin;LI Xiaohui(Department of Medical Imaging,The Second Affiliated Hospital of Xi’an Jiaotong University,Xi’an Shaanxi 710004,China)
出处
《中国医疗设备》
2022年第5期90-94,99,共6页
China Medical Devices
基金
陕西省自然科学基金(2022JQ-939)
陕西省重点研发计划一般项目(2021SF-267)。
关键词
低剂量CT扫描
深度学习图像重建算法
滤波反投影法
自适应统计迭代重建算法
low-dose CT scanning
deep learning image reconstruction algorithm
filtered back projection
adaptive statistical iterative reconstruction-Veo