Three-dimensional quantitative x-ray phase imaging

x射线相衬成像（XPCI）是近年来在x射线科学中出现的最令人兴奋的新方法之一。它基于折射和干涉现象而不是x射线衰减产生图像对比度，从而增强了图像中所有细节的可见性。此外，传统上认为“x射线不可见”的特征可以通过XPCI检测到。这在许多应用中具有变革性的力量，从医学到工业测试，通过生物学、文化遗产、材料科学、安全检查和许多其他领域。值得记住的是，无论是在科学领域还是在社会领域，x射线的使用都是无处不在的，所有使用x射线成像的领域都可以从XPCI中受益。几年前的问题是，XPCI被认为仅限于大型、专业和昂贵的设备，称为同步加速器——世界上只有大约50个同步加速器。然而，我的研究小组最近通过开发一种方法解决了这个问题，该方法使XPCI能够使用传统的x射线源（如医院中使用的x射线源）进行。这将允许将XPCI从超专业的实验室带到“现实世界”的应用中，并且与各种公司的谈判确实正在进行中，以使该技术进入商业开发。这个项目旨在开发下一代的这项技术。目前，我们的XPCI方法仅适用于二维“平面”成像应用。虽然这本身是有用的，并且在某些领域（例如，乳房x光检查或机场的行李扫描）有效地使用，但许多其他应用需要对成像样本进行全3D（“断层扫描”）重建。这是医学上一个众所周知的问题，例如，有些疾病不能用简单的“x光”诊断，而需要CT（计算机断层扫描）扫描；同样的原则也适用于许多其他领域，在这些领域中，样本的完整3D知识对于随后的决策过程至关重要。例如，在新药开发中，通常通过测试小动物的高分辨率3D图像来评估其效果，或者在测试复杂机械部件或新的“复合”材料时。因此，该项目旨在开发定量的、完整的3D版本的XPCI方法。这需要克服许多障碍，其中一些障碍具有非常技术性。例如，为了使x射线成像系统对x射线相位敏感，我们使用掩模，它覆盖了被成像物体的一部分。虽然这不会在平面成像中产生问题，因为被覆盖的样本部分比成像系统可以解决的最小元素（探测器像素）要小，但由于被称为欠采样的问题，在重建3D体积时确实会导致明显的伪影。这在其他学科（例如核医学）中也遇到过，研究人员已经开发出新的、更复杂的重建工具，可以解决或至少减轻这个问题。因此，我们计划使这些新的重建工具适应我们的XPCI方法的特定要求，以便可以进行可靠和定量的3D“相位”重建。最初，这将基于一个广泛的模拟阶段，在此期间，不同的算法将在不同的数据集上进行测试，这将能够确定最有希望的算法。接下来是一个实验阶段，我们将在真实的实验数据上测试算法：这将允许选择最佳解决方案并对其进行微调。最后，将有一个演示阶段，优化的3D方法将应用于实际的科学问题，其中包括关节软骨小损伤的3D可视化（众所周知，传统的x射线方法是看不见的），或者新一代复合材料的侵入/缺陷。

X-Ray Phase Contrast Imaging (XPCI) is one of the most exciting new methods emerged in x-ray science over recent years. It generates image contrast based on refraction and interference phenomena rather than x-ray attenuation, which enhances the visibility of all details in an image. Moreover, features classically considered "x-ray invisible" can be detected by XPCI. This has transformative power in many applications, from medicine to industrial testing, through biology, cultural heritage, material science, security inspections, and many other fields. It is worth remembering that the use of x-rays is all pervasive, both in science and in society, and all areas where x-ray imaging is used can strongly benefit from XPCI.The problem up to a few years ago was that XPCI was considered restricted to large, specialized and expensive facilities called synchrotrons - only approximately 50 of which exist in the world. However, my research group has recently solved this problem by developing a method that enables XPCI to be performed with conventional x-ray sources, like those used in hospitals. This will allow taking XPCI out of ultra-specialized labs and into "real-world" applications, and negotiations with various companies are indeed underway to take the technology into commercial exploitation.This project aims at developing the next generation of this technology. At the moment, our XPCI method works only in 2D, "planar" imaging applications. Although this is useful in itself, and is effectively employed in some areas (e.g. mammography or baggage scanning at airports), many other applications require the full 3D ("tomographic") reconstruction of the imaged sample. This is a well known problem in medicine, where for example some diseases cannot be diagnosed with a simple "x-ray" but require a CT (computed tomography) scan; the same principle also applies to many other areas, where full 3D knowledge of the sample is essential to the decision-making process that follows. Examples are in the development of new drugs, the effect of which is often assessed through high-resolution 3D images of the small animals on which they are tested, or in the testing of sophisticated mechanical parts or of new "composite" materials.This project therefore aims at the development of a quantitative, full 3D version of our XPCI method. This requires overcoming a number of obstacles, some of which have a very technical nature. For example, in order to make x-ray imaging systems sensitive to x-ray phase, we use masks, which cover parts of the imaged object. Although this does not create a problem in planar imaging, because the portions of the sample which are covered are smaller than the smallest element the imaging system can resolve (the detector pixel), it does result in significant artifacts when a 3D volume is reconstructed, because of a problem known as undersampling. This is also encountered in other disciplines (for example nuclear medicine), and researchers have developed new, more sophisticated reconstruction tools which allow solving or at least mitigating this problem. We therefore plan to adapt these new reconstruction tools to the specific requirements of our XPCI method, so that reliable and quantitative 3D "phase" reconstruction can be performed.Initially, this will be based on an extensive simulation phase during which different algorithms will be tested on various datasets, which will enable identifying the most promising ones. This will be followed by an experimental phase in which we will test the algorithms on real experimental data: this will allow selecting the best solution and fine-tuning it. Finally, there will be a demonstration phase in which the optimized 3D method will be applied to real scientific problems, among which for example the 3D visualization of small damage in articular cartilage (notoriously invisible to conventional x-ray methods), or of intrusion/defects in new-generation composite materials.