生物发光成像引导放疗的新型三维重建算法及其预临床评估

In the field of radiotherapy, cone-beam computed tomography (CBCT) has been widely used to guide radiation delivery. However, because of the low soft tissue contrast, it is almost infeasible to guide focal radiation for the soft tissue target in small animal studies. Bioluminescence tomography (BLT) is a highly sensitive 3D optical imaging modality that offers the possibilities to non-invasively monitor physiological and pathological processes at cellular level or an in vivo environment. In this proposal, we introduce the BLT into preclinical radiation research, as a solution to improve the small soft tissue target localization. An accurate and efficient reconstruction algorithm is a key component to achieve this goal. Specifically, this project will provide a 3D BLT reconstruction algorithm which is expected to accurately recover the bioluminescence source power and the morphology of the soft tissue target, and further reduce the irradiation target volume uncertainty. .To accurately reconstruct the shape of a bioluminescence target, the high-order simplified spherical harmonics approximation model of the radiative transfer equation will be adopted as the base to precisely describe photon transport in biological tissue. The development of fast solvers and the accelerating scheme of the inverse model are essential for our application. In particular, we propose an inverse model based on a hybrid sparse regularization combining total variation with Lp-norm penalty which is especially suitable for the sparse source distribution and shape reconstruction. To solve the inverse model, the development of Lp-nonconvex optimization algorithms is proposed to obtain stable and accurate solution. To further improve the reconstruction accuracy, the anatomical information provided by CBCT will be incorporated to limit the reconstruction solution space and multi-spectral measurements will also be adopted to better quantify the photon transport in tissue than single wavelength measurement. We will further develop an evaluation index to assess the performance of the proposed algorithms. The proposed algorithm development is expected to significantly extend the BLT application in the field of small animal research, especially advancing the soft tissue guidance radiotherapy and assessing the effects of treatment..

锥形束CT已在图像引导的放疗中广泛应用，但对于低对比度的软组织目标锥形束CT也无法定位引导。生物发光断层成像是一种可在细胞和分子水平上对生物体生理病理变化进行在体监测的高灵敏成像模态。本项目拟将其引入预临床放疗研究，实现对软组织目标的准确探测定位，减少放疗不确定性。为实现该目标，能准确恢复光源(靶目标)强度和形态的高效三维重建算法是关键。为此，拟采用辐射传输方程的高阶近似模型描述光子在生物组织的传输，研究其数值求解和加速方案，在此基础上构建混合总变差和Lp稀疏正则化的重建模型，发展非凸优化算法以获得快速、稳定、准确的重建结果。为提高精度，重建中采用锥形束CT提供的结构信息指导光学参数分配和限制解的范围，并结合多光谱测量数据克服逆问题不适定性。此外，拟设计切合应用的量化指标便于算法评估。本项目有望扩展生物发光断层成像在预临床研究中的应用，为引导软组织目标放疗和疗效评估提供影像基础和技术手段。

本项目将生物发光断层成像引入预临床放疗研究中，以解决图像引导放疗中低对比度的软组织定位引导难题。为实现对软组织目标的准确探测定位，减少放疗不确定性，我们着重研究能准确重建光源的高效三维重建算法，提出了基于排序L1范数稀疏正则的重建算法，探索了Lp（0<p<1）正则化在BLT光源重建中的应用，着重研究基于L1/2正则化的重建性能，还提出一种基于SCAD非凸正则化模型的重建算法，此外还研究了迭代收缩可行区域框架对重建算法的影响，并在小动物放疗平台上评估了项目组所提出的迭代收缩框架下的稀疏重建算法在BLT引导放疗中的定位准确性。大量的仿真实验和多组仿体及小鼠实验结果表明，所提出算法可以保证小于1mm的中心定位准确率。为了进一步提高算法在形状拟合方面的表现，又开发了一种联合L1和Laplacian流形正则的重建算法，并提出了针对多光源目标分辨的混合聚类以及同步聚类算法实现多光源的自动辨识和分析。