Development of an In-Silico Research Framework for Accelerating the Translation of Quantitative Photon-Counting Spectral Imaging to the Clinic

个性化患者治疗和评估治疗反应都是可以从分子信息中受益匪浅的任务。SPECT和PET提供分子成像，但价格昂贵，空间分辨率相对较差，需要专门的放射性药物和设施。MRI可以提供一些分子信息，但这些扫描仪速度很慢，许多患者由于金属植入物、心脏起搏器或幽闭恐惧症而无法使用MRI机器。理想的分子成像可以从X射线图像中获得，因为这些系统速度快，提供出色的空间分辨率，适用于几乎所有的患者群体。不幸的是，传统的X射线机无法提供分子信息，提供差的软组织对比度，并提供显着的电离辐射剂量。所有这三个问题都在一种新的X射线成像技术中得到了解决，称为X射线光子计数光谱成像（x-CSI），它提供了与CT空间分辨率相当的MRI软组织对比度，并且只有五分之一的辐射剂量。然而，关于如何最好地利用x-CSI为患者造福，仍然存在许多重要的问题。最佳像素尺寸、传感器材料、信号校正方案等是什么？如何重建光谱数据？哪些临床应用将从增加的信息中获益最多？计算机模拟通常用于回答这些问题，然而，由于对来自短程物理过程的失真的更高灵敏度以及因此所需的更复杂的电子器件，x-CSI模拟比传统的x射线模拟明显更复杂。因此，目前还没有工具能够足够详细地模拟x-CSI扫描仪，以充分回答这些问题。该项目旨在通过以下方式纠正这一问题：1。扩展我们现有的模拟框架，以更好地模拟短程物理过程，降低x-CSI图像和新的电子提出纠正他们。我们还将添加3D图像重建和图像分析工具，以便可以模拟治疗癌症患者的成像任务2。使用完整的框架来优化三种不同癌症相关成像任务中每一种的x-CSI扫描仪，考虑到我们的肿瘤学家和放射科医生合作者确定的一系列不同癌症类型3。优化一台通用x-CSI扫描仪，以执行所有三项临床成像任务4。将每个成像任务中的通用扫描仪与针对该任务优化的扫描仪进行比较，量化任何性能差异这项工作将为一系列利益相关者提供即时和长期利益。通过量化每个临床成像任务的通用和任务优化扫描仪之间的性能差异，这项工作将能够确定通用扫描仪是否适用于肿瘤学，或者是否需要任务优化的x-CSI扫描仪。结合针对各种肿瘤学任务确定的优化x-CSI扫描仪设计，这些信息将为寻求调整其肿瘤学扫描仪的医疗保健制造商提供信息，并为医生提供所需的信息，以支持可能影响临床决策的专业扫描仪。从长远来看，发布模拟框架的说明将允许更多的研究人员参与x-CSI研究，提供一个低成本的来源替代物理x-CSI扫描仪，不受限制地访问它生成的数据，并能够在成像链的每个阶段精确地了解地面实况。因此，该项目将加速x-CSI从实验室到诊所，并确保以最大限度地提高患者从这项尖端技术中获益的方式进行转移。

Personalising patient treatments and assessing treatment response are both tasks which could benefit greatly from molecular information. SPECT and PET offer molecular imaging but are expensive, have relatively poor spatial resolution and require specialist radio-pharmaceuticals and facilities. MRI can provide some molecular information, but these scanners are slow and many patients are unable to use MRI machines due to metal implants, pacemakers or claustrophobia. Ideally molecular imaging could be obtained from x-ray images, as these systems are fast, offer excellent spatial resolution and are suitable for almost all patient populations. Unfortunately, conventional x-ray machines are unable to provide molecular information, offer poor soft tissue contrast and deliver significant ionising radiation doses. All three of these problems are addressed in a new x-ray imaging technology, known as x-ray photon counting spectral imaging (x-CSI), which provides MRI comparable soft tissue contrast with CT spatial resolution and only 1 fifth of the radiation dose.x-CSI technology is just now entering clinical trials, with all major healthcare manufacturers working on developing their own system. Yet many important questions remain regarding how x-CSI can best be exploited for patient benefit. What are the best pixel sizes, sensor materials, signal correction schemes etc.? How should the spectral data be reconstructed? What clinical applications would benefit most from the added information? Computer simulations are normally used to answer these questions, however x-CSI simulations are significantly more complicated than conventional x-ray simulations due to the higher sensitivity to distortions from short range physics processes and consequently the more complicated electronics required. There are thus currently no tools capable of modelling an x-CSI scanner in enough detail to answer these questions fully. This project seeks to redress this by:1. Extending our existing simulation framework to better model short range physics processes that degrade x-CSI images and the novel electronics proposed to correct for them. We would also add 3D image reconstruction and image analysis tools so that imaging tasks used in treating cancer patients can be simulated2. Using the completed framework to optimise an x-CSI scanner for each of three different cancer related imaging tasks, considering a range of different cancer types as identified by our oncologist and radiologist collaborators3. Optimising a single general-purpose x-CSI scanner for performing all three clinical imaging tasks4. Comparing the general-purpose scanner in each imaging task with the scanner optimised for that task, quantifying any performance differencesThis work would provide both immediate and longer-term benefits to a range of stakeholders. By quantifying performance differences between a general-purpose and task optimised scanner for each clinical imaging task, this work will be able to determine whether a general-purpose scanner will be suitable in oncology, or whether task optimised x-CSI scanners are necessary. Combined with the optimised x-CSI scanner designs determined for the various oncology tasks, this information will both inform healthcare manufacturers seeking to adapt their scanners for oncology, and empower doctors with the information needed to argue for specialist scanners where these could affect clinical decisions. Longer term, publishing instructions for the simulation framework will allow more researchers to engage in x-CSI research by providing a low-cost source alternative to having a physical x-CSI scanner, unrestricted access to the data it generates and the ability to know the ground truth precisely at each stage of the imaging chain.This project would thus accelerate the translation of the x-CSI from the lab to the clinic and ensure that transfer occurs in a way which maximises patient benefit from this cutting-edge technology.