基于高阶统计形变模型的非刚体脑图谱自动配准

Recently， brain atlas registration is one of the popular research in the field of medical image processing. It is the key step of monitoring and diagnosis of disease, the brain region positioning, making surgical planning, the brain development research and etc. Usually, brain atlas registration uses non-rigid registration methods to warp the brain atlas to the patient brain image ,which the map information can be transferred to the image of the patient and provide tremendous help to the doctor. Or Warping the patient image to the brain atlas space so as to study the brains of different sizes and different shapes. Now, the main research has focused on the registration accuracy and the speed of registration. In our study, the main research is as follows: (1)constructing high order statistics deformation model, (2) registration based on higher order statistics deformation model, (3) registration accuracy evaluation, (4) the application of the developed algorithm to the segmentation of the brain regions and Alzheimer's disease.

脑图谱配准是近年来在医学图像处理领域的热门研究方向之一， 它是监测与诊断疾病，脑区定位，制定手术计划，大脑发育研究等的基础步骤。脑图谱配准主要采用非刚体配准的方法，将脑图谱和患者脑图像配准，图谱的信息就能在病人的图像中显示,给医生提供巨大的帮助。 或将患者的图像通过配准归一化到图谱空间， 有利于对不同大小和不同形状的大脑进行研究。目前主要研究集中在提高配准精度和配准速度。本课题围绕上述方面展开研究，有关内容如下：（1）高阶统计形变模型的建立，(2)基于高阶统计形变模型的配准，（3）配准精度评价，（4）将所开发的算法应用在脑区分割和老年痴呆症研究。

围绕医学图像的配准，分割，重建等问题展开研究，取得了一些有特色的创新性成果。.本研究完成了高阶形变模型的建立以及基于高阶形变模型的配准，并利用相关研究对老年痴呆病的诊断进行了探讨。同时利用变形优化从图像中重建三维血管，以及采用levelset对医学图像分割展开了研究，取得较好的结果。