小动物医学成像的快速发展为自动化影像分析提出了迫切需求。数字化小动物图谱与个体影像的配准是自动化图像分析的重要方法，但是不同动物个体之间存在着形态上的差异，不同的影像模式也有着各自的图像特征，这两个问题给图谱的自动精确配准带来了困难。本课题旨在研究一种小鼠全身数字图谱与不同小鼠影像模式之间的自动配准算法，采用近期构建的可变形小鼠全身图谱作为模板，通过优化图谱的骨骼姿态、身长、体重等参数来弥补图谱与个体之间的形态差异。为实现图谱与CT、MRI、PET、SPECT、光学照片、X光投影、体表扫描影像等模式的自动配准，本研究将不同影像模式按照空间维度分类，针对每一种维度设计配准方法，用低维度配准结果作为高维度配准的初始化条件以增强全身姿态配准的鲁棒性，并用低维度的图谱配准结果作为先验知识，辅助对三维图像中高对比器官的自动分割。本研究将为生物医学领域的小动物影像分析提供有利的算法支持。

Due to the fast development of high-throughput small animal imaging techniques, automated analysis of small animal images has become an urgent need. Automated registration of digital animal atlas to individual image is a widely-used method for small animal image analysis. However, because of the anatomical differences between individual subjects and the variety of image features between different imaging modalities, accurate registration of the animal atlas remains a challenging problem. This study aims at developing an automated atlas registration method for various preclinical mice imaging modalities, it will use a recently developed deformable mouse atlas to compensate for the anatomical differences of body posture, size and weight. To register this atlas to various modalities including CT, MRI, PET, SPECT, optical imaging, X-ray projection and surface scan, we classify these modalities according to their spatial dimension, and design dedicated registration scheme for each dimension. Moreover, atlas registration results of lower dimension will be used to initialize the registration of higher dimension. We also propose a method for automatically segmenting the high-contrast organs from three dimensional mice images using the registered atlas as prior information. This study will provide algorithm support for preclinical small animal image analysis.