ISSN 1004-4140
CN 11-3017/P
陈宗桂, 魏宁宁, 李舟帆, 等. 基于先验图像对颅内金属伪影的校正研究[J]. CT理论与应用研究(中英文), xxxx, x(x): 1-9. DOI: 10.15953/j.ctta.2024.002.
引用本文: 陈宗桂, 魏宁宁, 李舟帆, 等. 基于先验图像对颅内金属伪影的校正研究[J]. CT理论与应用研究(中英文), xxxx, x(x): 1-9. DOI: 10.15953/j.ctta.2024.002.
CHEN Z G, WEI N N, LI Z F, et al. Methods Research on Correction for Intracranial Metal Artifacts Based on Prior Image Utilization[J]. CT Theory and Applications, xxxx, x(x): 1-9. DOI: 10.15953/j.ctta.2024.002. (in Chinese).
Citation: CHEN Z G, WEI N N, LI Z F, et al. Methods Research on Correction for Intracranial Metal Artifacts Based on Prior Image Utilization[J]. CT Theory and Applications, xxxx, x(x): 1-9. DOI: 10.15953/j.ctta.2024.002. (in Chinese).

基于先验图像对颅内金属伪影的校正研究

Methods Research on Correction for Intracranial Metal Artifacts Based on Prior Image Utilization

  • 摘要: 目的:为了有效抑制颅内金属伪影对术后复查的影响,提出一种基于先验图像的金属伪影校正算法。方法:首先对颅内金属伪影的CT图像进行自适应滤波、阈值分割和K均值聚类获得只含金属物质的图像和先验图像;以先验图像为参照,在金属物质的投影轨迹中,对原始图像的投影数据进行插值校正得到修复后的图像,并将其与只含金属物质图像融合,获得最终校正图像。为了验证该算法抑制CT金属伪影的有效性,利用仿真单金属伪影和颅内金属伪影的CT图像进行伪影校正。结果:与线性插值算法和ADN算法相比,该算法校正后的图像结构相似性和峰值信噪比最大,而均方根误差最小。由两位高年资医生对该算法去除CT金属伪影的主观评价得分高于线性插值算法和ADN算法的主观评价得分。结论:该算法在有效保留图像边缘信息的同时,可有效地抑制CT金属伪影。

     

    Abstract: Objective: We propose a method to effectively suppress the effect of intracranial metal artifacts on postoperative review, a metal artifact correction algorithm based on prior imaging. Method: The CT image compensating for the intracranial metal artifact is obtained by adaptive filtering, threshold segmentation, and K mean clustering. In projecting the trajectory of metallic matter, the interpolation correction of the original image's projection data is obtained using the prior image as a reference, and the image is fused with the metal-containing material only image to produce the final corrected image. To verify the efficacy of this algorithm in suppressing CT metal artifacts, artifact correction was performed using CT images of simulated isolated metal artifacts and intracranial metal artifacts. Results: Compared with linear interpolation and the ADN algorithms, this algorithm achieves the highest structure similarity and peak signal-to-noise ratio and lowest root mean square error in corrected images. Two senior doctors subjectively scored this algorithm higher than the linear interpolation and ADN algorithms for removing CT metal artifacts. Conclusion: This algorithm effectively suppresses metal artifacts while preserving image edge information.

     

/

返回文章
返回