能谱CT虚拟单能图像重建和金属伪影去除算法在减少脊柱金属植入物伪影中的应用

Virtual monoenergetic image reconstructions and metal artifact reduction algorithms in spectral CT for reducing artifacts of spinal metal implants

目的 比较能谱CT金属伪影去除算法及虚拟单能图像重建与传统迭代重建在减少脊柱金属植入物伪影的差异。方法 56例脊柱矫形术接受金属植入物行标准能谱CT检查,包括常规迭代重建、金属伪影去除算法和虚拟单能图像重建。测量衰减系数(HU)和噪声(SD),以计算椎旁肌和椎管的信噪比。两名放射科医师独立评价图像质量和伪影减少程度。结果 与常规迭代重建相比,金属伪影去除算法和高keV虚拟单能图像显著降低低密度伪影及高密度伪影。与常规迭代重建相比,金属伪影去除算法和高keV虚拟单能图像椎旁肌(34.6±17.0HU vs. 26.1±13.5HU及34.6±17.0HU vs. 27.0±14.2)和椎管(102.5±60.1HU vs. 72.1±39.3HU及102.5±60.1HU vs. 60.1±38.0HU, P 均<0.05)的噪声伪影减少。观察者间评价主观图像质量的一致性良好,ICC=0.74。在主观图像质量评价中,金属伪影去除算法和高keV虚拟单能图像上表现出伪影减少分别为44/56例(78.6%)、48/56例(85.7%)。结论 能谱CT金属伪影去除算法和高keV虚拟单能图像重建上客观及主观伪像均减少,金属伪影去除算法联合虚拟单能图像的组合可能有希望进一步减少伪影。

Objective To detect the artifact reduction of spinal implants using metal artifact reduction algorithms (MAR), virtual monoenergetic image reconstructions and traditional iterative reconstruction in spectral detector CT. Methods Fifty-six patients undergoing spinal orthopedic surgery underwent standard-spectrum CT examinations, including routine iterative reconstruction, MAR and virtual monoenergetic image reconstructions. Attenuation (HU) and noise (SD) were measured to calculate signal-to-noise ratio of paravertebral muscles and spinal canal. Two radiologists independently assessed image quality and artifact reduction. Results MAR and high-keV virtual monoenergetic image reconstructions significantly reduced low-and high-density artifacts compared to conventional iterative reconstructions. Compared with the conventional iterative reconstruction, the noise artifacts of paraspinal muscles (34.6±17.0) HU vs.(26.1±13.5) HU and (34.6±17.0) HU vs.(27.0±14.2) and spinal canal (102.5±60.1) HU vs.(72.1±39.3) HU and (102.5±60.1) HU vs.(60.1±38.0) HU, all ( P <0.05) were reduced in MAR and high-keV virtual monoenergetic image reconstructions. Raters showed a good interrater agreement (ICC=0.74). For subjective image quality assessment, MAR and high-keV virtual monoenergetic image reconstructions yielded an artifact reduction in up to 44/56(78.6%) and 48/56 cases (85.7%), respectively. Conclusion Objective and subjective artifacts are reduced on both MAR and high-keV virtual monoenergetic image reconstructions in spectral detector CT, tHUs, the combination of the two algorithms may hopefully further reduce artifacts.