ISSN 1004-4140
CN 11-3017/P

不同迭代重建算法在眼眶CT中的适用性研究

Applicability of Different Iterative Reconstruction Algorithms in Orbital Computed Tomography

  • 摘要: 目的:探讨不同迭代重建算法在眼眶CT中的应用价值。方法:回顾性收集2024年1月至2024年3月首都医科大学附属北京同仁医院行眼眶CT检查的31例患者,分别使用混合迭代重建算法(iDose4)和全模型迭代重建算法(IMR)技术对图像进行标准算法和骨算法重建,对比分析不同重建算法图像的平均CT值、噪声(SD)、信噪比(SNR)、对比度噪声比(CNR)以及图像质量主观评分;采用随机区组方差分析比较不同算法重建图像间的客观指标差异;采用多个相关样本的秩和检验比较不同算法重建图像的主观评分差异。结果:不同标准算法重建图像的CT值、噪声、SNR、CNR差异均有统计学意义,与iDose4相比,IMR标准算法重建图像噪声降幅达25%~67%,SNR增加1.3~1.5倍,CNR增加2~3倍;与iDose4相比,IMR骨算法重建图像噪声降幅达70%~96%,SNR增加5~15倍,CNR增加4~31倍;不同骨算法及标准算法的图像质量评分差异均有统计学意义,标准算法IMR1-Brain Routine组评分最高5(4,5),骨算法iDose43 YD组评分最高5(5,5);两名观察者评分的一致性较强,加权Kappa系数0.644~1。结论:眼眶CT中推荐使用混合迭代重建技术iDose4 Level 3重建骨算法图像,使用全模型迭代重建技术IMR Level1重建标准算法图像。

     

    Abstract: Objective: This study aimed to investigate the value of different iterative reconstruction algorithms for orbital computed tomography (CT). Methods: Orbital CT data from 31 patients were retrospectively collected from January to March 2024 at Beijing Tongren Hospital Capital Medical University. The images were reconstructed using both the Hybrid Iterative Reconstruction Algorithm (iDose4) and Iterative Model Reconstruction (IMR) techniques for the bone and standard algorithms. The average CT number, standard deviation (SD) of noise , signal-to-noise ratio (SNR), contrast-to-noise ratio (CNR), and image quality scores of different reconstruction algorithms were compared. Randomized block variance analysis was used to compare differences in objective indicators, and the Wilcoxon signed-rank test for related samples was used to compare differences in subjective scores. Results: The differences in the CT number, noise, SNR, and CNR of images reconstructed using different standard algorithms were statistically significant. Compared with iDose4, the noise reduction of the IMR standard algorithm-reconstructed images ranged from 25% to 67%, with an SNR increase of 1.3~1.5 times, and a CNR increase of 2 to 3 times. Compared with iDose4, the noise reduction of the IMR bone algorithm-reconstructed images ranged from 70% to 96%, with an SNR increase of 5~15 times and a CNR increase of 4~31 times. The differences in the image quality scores of the images reconstructed using different bone and standard algorithms were statistically significant. The highest scores were 5 (5, 5) for the iDose4 Level 3 Y-Detail (YD) bone algorithm group and 5 (4, 5) for the IMR Level 1 Brain Routine standard algorithm group. The consistency of the scores evaluated by the two observers was strong, with weighted kappa coefficients ranging from 0.644 to 1. Conclusion: In orbital CT, we recommend using a hybrid iterative reconstruction technique (iDose4 Level 3) to reconstruct bone algorithm images and the iterative model reconstruction technique (IMR-Level 1) to reconstruct standard algorithm images.

     

/

返回文章
返回