【作者】 张丹枫；

【导师】 王志中； 赵俊；

【作者基本信息】 上海交通大学 ， 生物医学工程， 2011， 博士

【摘要】 提高胃结肠癌症早期诊断的准确性和效率具有重要意义。目前以虚拟内窥镜为代表的一系列非侵入性的检查方法已逐渐成为胃肠道息肉检查和癌症早期诊断的常用技术。其中,近年来新提出的虚拟外翻技术能够通过外视的视角观察原内壁表面信息。观察经虚拟外翻后的结肠数据,视野明显扩大,视野面积数倍于虚拟内窥镜,而对病灶的定位也更为直观。这些特点克服了原有虚拟内窥镜和虚拟展平技术的局限性,使得虚拟外翻技术具有重要的实用价值。其应用前景经专家论证,已获得了国家863项目基金资助。然而虚拟外翻技术作为一项全新的虚拟可视化技术,还存在一些问题亟待完善以便更好地应用到临床实践。为了改进虚拟外翻算法的处理结果,本文提出了以完整体数据作为外翻对象的虚拟外翻算法。经模型数据和人体CT扫描数据上测试,体数据外翻结果明显改善了原有二值化数据外翻结果中数据表面粗糙不光滑或者结构破裂的现象,进而可以提高胃结肠息肉检查的准确率和效率,降低检查者的视觉疲劳。围绕体数据外翻算法,改进和优化了与体数据外翻相关的外翻基准面、中心路径、体数据采样网格等外翻关键材料的处理算法,并验证了各算法的有效性,分析了它们在胃和肠道中的适用特性。基于测地距离变换设计的局部膨胀腐蚀算法能够很好地避免原有膨胀腐蚀算法造成的基准面粘连问题;而利用了混合电场模型的虚拟外翻方法较原有基于局部电场模型的外翻方法大大减少了采样网格的计算消耗。通过分析各数据采样网格模型和外翻算法,本文总结出:通过中心路径外翻算法和混合电场模型采样网格更适合结肠内壁的虚拟外翻,而基准面外翻算法和符号距离函数采样网格更适合胃部内壁的虚拟外翻。关于外翻数据的检查方法,本文基于数据采样网格设计了两种专用于虚拟外翻数据检查的导航路径:平行导航线以及螺旋导航线。设计并研究了螺旋导航检查、定点检查、分段检查和多导航路径并行检查等多种能够遍历整个结肠外翻数据表面的导航检查模式,分析了检查模式中各参数之间的定量关系,推算出螺旋导航检查和定点检查中推荐的参数设定,为虚拟外翻数据应用于临床实践打好基础。关于外翻数据检查效果评估,利用体数据外翻方法对模型数据和真实CT数据实施外翻后,将虚拟外翻数据的息肉检查和虚拟内窥镜息肉检查进行了对照实验比较。结果表明无论在模型数据还是真实CT数据上,虚拟外翻数据检查都比虚拟内窥镜方法大幅降低了检查时间(在模型中平均减少了65.0%,在CT数据中平均减少了72.1%)。同时基本保持了对息肉检查的敏感性、特异性水平。虚拟外翻技术可以作为一种补充的非侵入式技术应用于胃肠道息肉检查。甚至和虚拟内窥镜、虚拟展平技术等融合在一起,组成功能更完整的胃肠道癌症诊断系统。更多还原

【Abstract】 To develop the observation techniques of stomach and colon will result in an increase in the diagnostic availability of the stomach and colon cancer. Many visualization methods for non-invasive gastrointestinal (GI) tract diagnosis been proposed. Some of which have been developed and adopted in clinical applications, such as virtual colonoscopy and virtual flattening. The virtual eversion method allows the observation of the colonic mucosa under a global view while keeping the colon path preserved. Reviewing the everted data, the view angle is fairly expanded, the locations of the lesions are more intuitive. All of these make the virtual eversion overcome the limitations in the virtual colonoscopy and virtual flattening.However, as a developing visualization methods, the following issues still need to be addressed before virtual eversion can be used in clinical applications. The virtual eversion method for GI volume data was proposed to perfect the eversion result. The volume data based eversion method resulted a much smoother surface, compared with the old binary data based eversion method. The smooth result may improve the accuracy of polyps detection. A set of key material for volume data eversion were presented in this paper, e.g. eversion reference surface, centerline, volume data sampling grid. The relevant algorithms were validated and analyzed.The proposed local dilation and erosion algorithm, which based on the geodesic distance transformation, resulted a perfect eversion reference surface without error connection in it. The data sampling grid based on hybrid electrical field model saved much processing time that cost by the old data sampling grid. The analytics on various data sampling grids and eversion methods resulted that, the eversion method based on centerline model and hybrid electrical field model was suitable for colon data eversion, and the method based on reference surface and signed distance function model was good for stomach eversion.We designed two kinds of navigation path based on the data sampling grid, the parallel navigation path and the spiral navigation path, for observing the everted data. Several observation schemas (spiral navigation, check-points observation, section observation, and multiple path navigation) were researched. We analyzed the dependency of the parameters in the observation schemas, and figured out the suggested parameters. The schemas and parameters are important foundation for the everted data observation.The volume eversion method were validated using 3D phantoms and CT data sets. A study on the observation performance of the everted data showed that, the reading times were (65.0% of time reduction for Phantoms, and 72.1% of time reduction for CT data) less than those using virtual colonoscopy, while maintaining the sensibility and specificity. The virtual eversion can be a complimentary method for non-invasive polyp inspection. Together with virtual colonoscopy and virtual flattening, the virtual eversion may be integrated to produce a powerful system for GI tract diagnosis.更多还原