基于PET/CT探测环真光子数呼吸运动门控的研究

Research for Respiratory Motion Gated with Detector Ring True Coincidence Event of PET / CT

呼吸运动是导致PET/CT胸腹部成像质量下降的主要原因,探寻一种行之有效的方法来降低呼吸运动对PET/CT成像质量的影响,在临床上对疾病的诊疗显得尤其重要。提出一种基于探测环真光子数的呼吸运动门控方法来提高PET/CT胸腹部诊断图像质量,利用PET探测环真光子数分布会随着体模的运动而发生相应变化的特征来对原始数据进行呼吸运动门控处理。采用GATE软件仿真PET/CT成像的过程,分别使用几何体模和像素体模(NCAT)仿真肺部在PET/CT扫描过程中的运动。然后,使用提出的方法门控仿真扫描数据,并重建出三维图像。相对于运动模糊图像,门控图像目标区域的形状、大小和位置等方面更接近于静态图像。在冠状面上,门控图像质量优于运动模糊图像,其图像结构相似度(SSIM)分别提升5%、3%、9%;对于矢状面的line profiles,门控图像与静态图像接近,优于运动模糊图像。结果表明,所提出的方法能有效减少PET/CT图像的呼吸运动伪影,使肺部肿瘤的形状、大小和位置接近静态图像,并且有效地克服目前一些校正方法的局限性。

The main degrading factor in the quality of PET/CT image of the thorax and upper abdomen is respiratory motion, exploring an effective method to reduce this effect in clinical is especially important. To improve the image quality, this paper presents a new respiratory motion gating method based on the detector ring true coincidence event, which use the feature of detector ring true coincidence event distribution changing with the movement phantom to gating the raw data. GATE （Geant4 Application Tomographic Emission） software package was used to simulate realistic PET/CT scanner, and the geometry phantom and NCAT （NURBs Cardiac Torso） phantom were built to simulate the lung motion. Next, the simulation raw data were gated by the proposed method and the 3 - D image was reconstructed. Compared with the motion-blurred image, the shape, size and position of the gated image is closer to the static image. In the coronal image, the quality of the gated image is better than the motion-blurred image; the structural similarity （SSIM） were respectively raised 5 % , 3% and 9% in the shape, size and position; in the sagittal image, the line profile curve of the gating image is similar to the static image, which is better than the motion-blurred image. These results demonstrated that the proposed method reduced the motion artifact of PET/CT image effectively, making the lung cancer image more accurate, furthermore, effectively overcome the limitations of some methods.