Novel image reconstruction techniques with application to proton radiotherapy for optimisation of cancer treatment

新颖的图像重建技术应用于质子放射治疗以优化癌症治疗

• 批准号: EP/R009988/1
• 负责人: Nikolaos Dikaios
• 金额: $ 12.88万
• 依托单位: University of Surrey
• 依托单位国家: 英国
• 项目类别: Research Grant
• 财政年份: 2018
• 资助国家: 英国
• 起止时间: 2018 至 无数据
• 项目状态: 已结题
• 来源: https://gtr.ukri.org/projects?ref=EP%2FR009988%2F1
• 关键词: Novel; image; reconstruction; techniques; application

One in two people will develop cancer, and cancer is the cause of approximately one third of all UK deaths. Radiotherapy accounts for >40% of curative treatments, owing its effectiveness in the ability to accurately target tumours. Proton Beam Therapy (PBT) is rapidly gaining momentum compared to x-ray/electron beams with more than 63 operating sites and 40 sites under construction worldwide. Protons have similar relative biological effectiveness (RBE) to photons, but an excellent depth-dose distribution profile that allows better conformation of dose distribution to target compared to x-rays or electrons, thereby reducing the integral dose to the body and avoiding to dose normal tissue structures near the tumour. This is crucial when treating growing children, to avoid side effects such as developmental delays, hormone deficiencies, effects on bone and muscle tissue, and hearing loss or damage to salivary glands. The two main methods to deliver proton beams is via passive spreading and spot scanning PBT. Spot scanning PBT is a new therapy that penetrates deeper and produces fewer neutrons than passive spreading, further decreasing the integral dose and the risk of secondary cancer, but is more sensitive to spatial errors increasing the risk of delivering the dose in the wrong place. Spot scanning PBT can deliver treatments in sub-mm accuracy, but because of imaging limitations prior to treatment, upon which the treatment is planned it currently cannot achieve more than 7 mm accuracy. To maximise the potential of PBT it is crucial to accurately know the dose distribution and be able to shape and control it, making imaging the number one challenge for accurate treatment planning. Based on the reconstructed images, proton stopping power maps are calculated, which inform us about the dose distribution. Due to the proton Bragg peak characteristic a miscalculation in the proton stopping power map (distance from the 90% to the 10% dose level is only a few mm) can result in the proton beam missing its target and damaging healthy tissue, while the tumour receives much lower dose. This work will initially quantify proton stopping power maps directly from proton CT (pCT), aiming to reduce range uncertainties and enhance PBT accuracy. pCT measures the energy loss for protons traveling along tracks allowing the estimation of the integrated relative electron density with respect to a reference medium along the proton path. Proton stopping power maps can be estimated directly by inverting the path integral. Current reconstruction algorithms, such as the filter back-projection approaches, falsely assume Gaussian energy straggling distribution, and do not account for multiple Coulomb scatter (MCS). Reconstructed pCT images become blurred by MCS, which results in a resolution of around 3-5 mm and the energy spread distributions in fact resemble asymmetric Gaussian functions due to electronic energy-loss straggling and MCS. The proposed pCT reconstruction will account for non-Gaussian energy loss distributions, and iteratively correct for MCS to improve the spatial resolution and accuracy of proton stopping power maps. With expected anatomical changes of both tumour and normal tissue during a typical 5-7 week course of radiation, relying solely on a pCT acquired before therapy will lead to under dosing of the tumour and/or unnecessary exposure of organs at risk to higher doses. This proposal addresses ways to fuse information from pCT (acquired before treatment) in the reconstruction of limited number of projections during spot scanning PBT, and update the proton stopping power map during the treatment. The proposed methods will make on-treatment imaging feasible, allowing for significant improvement in treatment planning.

每两个人中就有一个会患上癌症，而癌症是英国大约三分之一死亡的原因。放射疗法占治愈性治疗的40%以上，这是由于其能够准确靶向肿瘤的有效性。与X射线/电子束相比，质子束治疗（PBT）正在迅速发展，全球有超过63个运营地点和40个正在建设中的地点。质子具有与光子相似的相对生物学有效性（RBE），但是具有优异的深度-剂量分布轮廓，与X射线或电子相比，其允许更好地将剂量分布构造到目标，从而减少对身体的积分剂量并避免对肿瘤附近的正常组织结构进行剂量。在治疗成长中的儿童时，这是至关重要的，以避免副作用，如发育迟缓，激素缺乏，对骨骼和肌肉组织的影响，听力损失或唾液腺损伤。质子束的传输主要有两种方法：被动扩散和点扫描PBT。点扫描PBT是一种新的治疗方法，与被动扩散相比，穿透更深，产生更少的中子，进一步降低了积分剂量和继发性癌症的风险，但对空间误差更敏感，增加了在错误位置提供剂量的风险。点扫描PBT可以提供亚毫米精度的治疗，但由于治疗前的成像限制，治疗计划目前无法达到超过7毫米的精度。为了最大限度地发挥PBT的潜力，准确了解剂量分布并能够对其进行成形和控制至关重要，这使得成像成为准确治疗计划的头号挑战。基于重建的图像，质子阻止本领地图计算，这告诉我们的剂量分布。由于质子布拉格峰特性，质子阻止功率图中的错误计算（从90%到10%剂量水平的距离仅为几mm）可能导致质子束错过其靶点并损伤健康组织，而肿瘤接受的剂量要低得多。这项工作最初将直接从质子CT（pCT）中量化质子阻止功率图，旨在减少范围不确定性并提高PBT准确性。pCT测量质子沿沿着轨道行进的能量损失，允许估计相对于沿质子路径沿着的参考介质的积分相对电子密度。目前的重建算法，如滤波反投影方法，错误地假设高斯能量离散分布，并没有考虑到多个库仑散射（MCS）。重建的pCT图像由于MCS而变得模糊，这导致约3-5 mm的分辨率，并且由于电子能量损失离散和MCS，能量扩散分布实际上类似于不对称高斯函数。所提出的pCT重建将考虑非高斯能量损失分布，并迭代地校正MCS以提高质子阻止本领图的空间分辨率和准确度。在典型的5-7周放射过程中，由于肿瘤和正常组织的预期解剖学变化，仅依赖于治疗前获得的pCT将导致肿瘤剂量不足和/或有风险的器官不必要地暴露于更高剂量。该提案提出了在点扫描PBT期间重建有限数量的投影时融合来自pCT（治疗前采集）的信息的方法，并在治疗期间更新质子阻止功率图。提出的方法将使治疗中成像变得可行，从而显着改善治疗计划。

项目成果

Proton Computed Tomography: A Case Study for Optimal Data Acquisition

质子计算机断层扫描：最佳数据采集案例研究

• DOI: 10.1109/nss/mic42101.2019.9060034
• 发表时间: 2019
• 作者: Panagiotidou M

Statistical limitations in ion imaging.

• DOI: 10.1088/1361-6560/abee57
• 发表时间: 2021-05-10
• 期刊: Physics in medicine and biology
• 影响因子: 3.5
• 作者: Collins-Fekete CA;Dikaios N;Bär E;Evans PM
• 通讯作者: Evans PM

Molière maximum likelihood proton path estimation approximated by cubic Bézier curve for scatter corrected proton CT reconstruction.

莫里哀最大似然质子路径估计通过三次贝塞尔曲线近似，用于散射校正质子 CT 重建。

• DOI: 10.1088/1361-6560/ab9413
• 发表时间: 2020
• 期刊: Physics in medicine and biology
• 影响因子: 3.5
• 作者: Lazos D

Integration of Proton Computed Tomography into the Open Source Software STIR

将质子计算机断层扫描集成到开源软件 STIR 中

• DOI: 10.1109/nss/mic42101.2019.9059915
• 发表时间: 2019
• 作者: Panagiotidou M

Machine learning for proton path tracking in proton computed tomography.

质子计算机断层扫描中质子路径跟踪的机器学习。

• DOI: 10.1088/1361-6560/abf1fd
• 发表时间: 2021
• 期刊: Physics in medicine and biology
• 影响因子: 3.5
• 作者: Lazos D