迭代算法对腹部CT图像影响及降低剂量的研究

Impact of Iterative Reconstruction Algorithm on Abdominal CT Image Quality and Radiation Dose

目的:探讨CT不同重建算法对不同层厚腹部图像噪声和图像质量的影响及迭代重建算法在降低腹部辐射剂量的潜能。方法:利用GE 64排能谱CT(Discovery CT750 HD)对68例病人前瞻性地行腹部CT平扫,将原始数据保存并分别用滤波反投影重组(FBP)、自适应统计迭代重组技术(ASiR)和基于模型的迭代重组技术(MBiR)三种重建算法重组成0.625mm、1.25mm、2.5mm和5mm不同层厚的CT图像。对图像进行定性和定量评估,并采用配对样本t检验比较。结果:对四种不同层厚腹部CT图像利用FBP、50%ASiR和MBiR三种重组算法的图像数据显示,MBiR重组算法图像噪声相对于50%ASiR和FBP重组图像要低(P<0.01);图像评分均分别明显好于相同层厚50%ASiR和FBP重组的图像;图像对比噪声比(CNR)均高于50%ASiR和FBP重组图像的CNR(P<0.01)。结论:MBiR重组的0.625mm层厚图像大大降低了图像噪声、提高了对比噪声比CNR,能代替传统的FBP重组的5mm图像,进一步提高了小病灶的检出率并能降低辐射剂量。

Purpose： To compare the image noise and image quality of abdominal CT images reconstructed with different slice thickness and different reconstruction algorithm techniques and to evaluate the potential of reducing radiation dose with iterative reconstruction algorithm. Methods： A total of 68 patients underwent plain abdominal CT with a 64-detector CT scanner （Discovery CT750 HD; GE Healthcare）. The projection data sets were reconstructed as images with 0.625mm, 1.25mm, 2.5 mm and 5ram thickness with FBP, ASiR and MBiR. Image quality, mean CT values, image noise and contrast-to-noise ratio （CNR） relative to the muscle and the liver with each algorithm were assessed. Paired t test was used for statistical analysis. Results： For images of different slice thickness, those images with MBiR reconstruction were with lower image noises when compared with those with 50%ASiR and FBP reconstruction （p〈0.01）. By using MBiR, the scores of image quality for 4 different slice thicknesses were significantly better than 50%ASiR and FI3P. CNR for the liver with MBiR was also much higher that with 50% ASiR and FBP. Conclusion： Advanced MBiR reconstruction algorithms can be used to reduce image noise and improves image CNR at 0.625mm slice thickness, and can be used to replace conventional FBP images with 5mm slice thickness. This may further improve the visualization of small lesion. MBiR techniques have the ability to reduce radiation dose through their improvement in image quality compared with available FBP, and can provide promising potentials for image quantitative analysis.