恶性肿瘤生物适形调强放疗PET/CT/MRI靶区智能勾画

Each year, the estimated number of new cancer cases in China is over 2 million and 70% of them need radiotherapy. In this era of high-precision biological conformal intensity-modulated radiotherapy, the accurate delieation of target volume with respect to tumor boundaries, shape, and volume is crucial, which may be achieved theoretically by the segmentation of PET tumor images. However, it is very difficult in practice due to the heterogeneity inside tumors and the low resolution , large noise and partial-volume effect(PVE) in PET tumor imaging. So, we propose that the PET, CT, and MRI textural characterizations such as coarseness, contrast, and busyness are extracted and used to the segmentation of PET tumor images via level set methodes and variational principles with clinical priors and a heterogeneous intensity model. The accuracy of the delineation of target volume is validated by pathologic examination. Moreover, the high-resolution image from CT or MRI is incorporated in the low-resolution PET image to yield a minimum PVE-corrected image and a maximum resolution PET reconstruction image by a maximum-a-posteriori approach.This project is very important to high-precision radiotherapy in theory and in clinical researches.

在每年新增的两百多万癌症患者中，70%以上需要进行放射治疗。生物适形调强放疗成功实施的关键在生物靶区的高精度勾画，其本质是肿瘤分子生物特性功能影像PET图像分割，但肿瘤内部生物特性的各向异性特性和目前临床肿瘤PET图像固有的低空间解析度、强噪声与严重的部分容积效应，使高精度的肿瘤放疗生物靶区图像分割非常困难。因此，本项目提出联合肿瘤PET和高解析度的CT、MRI高阶纹理特征，采用各向异性向量值变分图像分割模型和水平集方法，集成临床专家肿瘤靶区勾画和组织病理知识，进行肿瘤靶区图像的高精度分割和智能勾画验证评估临床研究。为进一步提高靶区分割精度，计划将高解析度的肿瘤CT、MRI解剖结构信息集成到PET图像的最大后验概率统计重建和部分容积效应校正模型中，进行最大化PET图像空间解析度的统计重建优化方法和最小化肿瘤PET部分容积效应校正方法研究。对恶性肿瘤高精度放疗具有重要理论和临床应用研究价值。

近年来，我国癌症患者不断增加，2015年我国新增癌症患者约430万，其中，60－70%的患者需要进行放射治疗。调强放疗成功实施的关键在于放疗靶区的高精度勾画，其本质是肿瘤影像PET/CT/MRI图像分割，但肿瘤内部生物特性的各向异性特性使高精度的肿瘤放疗靶区图像分割非常困难。.为攻克肿瘤放疗靶区高精度自动勾画难题，本项目组系统深入地开展了肿瘤靶区智能勾画研究，发现肿瘤PET、CT、MRI高阶纹理特征有利于区分肿瘤和正常组织器官，并据此发现，提出了一系列的靶区智能勾画新方法。进一步，提出了肿瘤生物调强放疗计划最优处方剂量方法和多目标优化方法。同时，项目组还开展了生物调强放疗疗效等临床应用研究。本项目研究成果对恶性肿瘤精准放疗具有重要理论和临床应用价值。.目前，本项目组已发表学术期刊论文10篇，国际顶级会议论文3篇，正在由专家评审的期刊论文5篇。设计、开发了肿瘤靶区智能勾画和调强放疗计划软件系统MBIG-IRT v1.0，已获得6项计算机软件著作权。主要成果如下：.1）PET/CT Image Textures for the Recognition of Tumors and Organs at Risk for Radiotherapy Treatment Planning；.2）Molecular Biological Characteristics Based Hierarchical Mumford-Shah Vector-Model for the Delineation of Biological Target Volumes Corresponding to Head and Neck Tumors；.3）Piecewise uniform dose prescription and optimization based on PET-CTimages；.4）头颈癌PET与MRI融合放疗靶区自适应区域生长勾画；.5）18F-FDG PET-CT功能影像引导下鼻咽癌同步加速生物调强放疗疗效分析；.6）头颈部肿瘤分子生物纹理分析与生物靶区自适应勾画；.7）头颈部肿瘤PET图像分割随机游走方法；.8）Positron Emission Tomography图像非局部几何非线性扩散去噪方法；.9）多方向几何非线性扩散图像去噪方法。