Modelling image formation and simplification of image aquisition for Talbot-Lau x-ray phase contrast CT

使用Talbot-Lau方法的相衬成像除了提供通常的吸收图像之外，还提供物体的微分相位图像和暗场图像。近10年来的研究表明，微分相位像和暗场像包含着与吸收像互补的有价值的信息。虽然该方法适用于普通X射线管，但多色光谱导致图像中的伪影。一个重要而特殊的方面与暗场有关。暗场可能是由颗粒状结构引起的，如肺泡或乳腺肿瘤中微米大小的钙化。这个非常有趣的贡献几乎可以完全被射束硬化或边缘的贡献所覆盖。此外，相位图像和吸收图像可能受到射束硬化的影响。应该注意的是，对于暗场，该效果可能是极端的：暗场图像可能仅包含由于射束硬化而导致的结构。这不会发生在相位和吸收图像中。Talbot-Lau方法需要获取所谓的相位步进曲线。该过程可以将用于CT成像的采集时间延长到可能引起不想要的伪影或甚至意味着不可忍受的采集时间的量。在应用项目中，将分别开发迭代CT重建策略。重要的是，吸收图像、微分相位图像和暗场图像这三个图像将不会被独立地重建。相反，通过对对象的物理特性进行建模，将考虑三个图像中的相关性，以便减少噪声并消除伪影。所应用的迭代重建主要旨在研究相关性，并且最初不旨在包括用于增强图像质量的其他最先进的措施。此外，它将研究是否可以省略相位步进CT成像，从而达到更容易的采集和更短的采集时间。甚至可以放弃参考相移曲线的采集。

Phase contrast imaging using the Talbot-Lau method delivers in addition to the usual absorption image the differential phase image and the darkfield image of an object. During the recent 10 years it has been proven that the differential phase image and the darkfield image contain valuable and complementary information to the absorption image. While the method is applicable at usual X-ray tubes, the polychromatic spectrum leads to artifacts in the images. An important and special aspect is related to the darkfield. The darkfield can be caused by granular structures like lung alveoli or micrometer-sized calzifications in breast tumors. This very interesting contribution can almost completely be overlayed by contributions from beam hardening or edges. Also the phase image and the absorption image may be affected by beam hardening. It should be noticed that the effect can be extreme for the darkfield: a darkfield image may exclusively contain structures due to beam hardening. This cannot occur in the phase and the absorption image. The Talbot-Lau method requires the acquisition of so-called phase stepping curves. This procedure can extend the acquisition time for CT imaging to an amount which may induce unwanted artifacts or even imply intolerable acquisition times. Within the applied project an iterative CT reconstruction strategy will be developed respectively further developed. The important ingredient is that the three images - absorption image, differential phase image and darkfield image - will not be reconstructed independently. Instead, by modeling the physics properties of the object the correlations in the three images will be taken into account in order to reduce noise and to eliminate artifacts. The applied iterative reconstruction primarily aims at the investigation of the correlations and is initially not intended to include other state-of-the-art measures for enhancing the image quality. Further, it will be investigated whether the phasestepping can be omitted for CT imaging thus reaching easier acquisition and shorter acquisition times. It might even be possible to waive the acquisition of the reference phasestepping curve.