面向肿瘤放疗决策的X射线发光计算机断层成像模型和算法研究

The development of image guided radiotherapy (IGR) technology, especially the application of cone-beam computed tomography (CBCT) has dramatically improved the accuracy of radiation delivery. However, it is almost infeasible to guide focal radiation for the soft tissue target with low contrast or early neoplastic lesions in small animal studies. X-ray luminescence computed tomography (XLCT) provides simultaneous structural imaging and molecular specificity imaging. Compared to other optical imaging modalities, XLCT has more advantages in detection depth and convenience of clinical transformation, which make it be a potential and powerful tool for image-guided radiotherapy. In this proposal, we introduce the XLCT into preclinical radiotherapy research, as a solution to improve the localization of small soft tissue target or small early neoplastic lesions, reducing uncertainty of radiation delivery. The key to obtain high quality images used for IGR is the development of fast and robust 3d reconstruction algorithm, which can accurately reconstruct target strength and morphology from sparse sampling with reduced X-ray exposure time and dose. Therefore, we intent to adopt hybrid transportation model to balance the modeling accuracy and computation efficiency, by combing the high-order and low-order approximation of the radiative transfer equation. To achieve accurate reconstruction sparse sampling, we intent to introduce graph model on finite element mesh and build up a novel reconstruction model that combines Graph Laplacian regularizer and sparse L1 regularizer, improve the shape-fitting ability of the reconstruction algorithm by jointing multiple meshes, develop efficient algorithms based on greedy strategy or convex optimization, introduce cluster strategy to yield automatic identification of multiple targets, and design practical quantitative indicator to assess the proposed method. This project is expected to provide new imaging basis and technological means for image guided radiotherapy and curative effect evaluation.

影像引导的放疗技术的开展，尤其是锥束CT（CBCT）的应用极大提高了放疗精度，但对软组织和早期肿瘤病变CBCT无法实现定位引导。X射线发光计算机断层成像（XLCT）可同时实现结构和特异性成像，相比其它光学成像，在探测深度和临床转化上更具优势，有望成为新型图像引导放疗有力工具。本项目拟将其引入预临床放疗研究，实现对软组织和早期肿瘤组织等的准确定位。为此，基于稀疏采样策略以减少X射线照射时间和剂量，发展准确恢复靶目标强度和形态的快速、鲁棒三维重建算法获得优质的重建图像是关键。拟采用兼顾准确性和计算效率的辐射传输方程高低阶近似描述光在生物组织内的传输，通过稀疏采样，在有限元上引入图模型，结合图拉普拉斯正则和稀疏正则构建全新重建模型，联合多网格信息提高算法定位和形状拟合能力，发展基于贪婪策略和凸优化的快速算法，并引入聚类策略自动辨识多目标，设计量化评估方法。本项目有望为影像引导放疗提供新成像方法。

放射治疗是目前临床治疗肿瘤的最主要手段之一，影像引导放疗技术对放疗中特异性和肿瘤早期治疗有重要的价值。X射线发光计算机断层成像既可以获得解剖结构像，又可以实现分子特异性成像，在影像引导放疗技术中有天然的优势。为提高X射线发光计算机断层成像的速度、精度以满足临床中的应用，本项目围绕准确高效的光传输模型、组织器官分割和精准高效的光源重建算法几个方面展开，研究辐射传输方程的各种近似模型及更加准确高效的自适应混合光传输模型，提出基于CoSaMP和SP3模型的荧光分子层析成像方法以及多光谱策略下的光传输模型；评估主动形状模型分割方法对光学重建的影响，提出了环状血管的单像素宽度中心线提取方法，基于多图谱配准和随机森林的微型CT肾脏分割方法，为后续光学段层成像方法的研究提供准确的结构模型数据；设计准确高效的XLCT重建算法，研究基于L1-2正则化的荧光分子段层成像方法、半阈值追踪算法等高效鲁棒的重建算法和策略在减少测量信息的同时提高了重建精度和成像质量。本项目为影像引导的精准放疗提供新的技术手段，拓展了光学分子断层成像在基础生命科学研究中的应用，也可为其它反问题求解提供借鉴，具有重要的理论价值和应用前景。