Hybrid non-rigid registration method for registering rat skeletons from micro-CT images

Objectives A hybrid non-rigid registration method for automatic rat spine detection and registration and further rat limbs registration was proposed.

Methods We first developed an automatic algorithm to extract the rat spine from its whole-body skeleton and then detect top point on the spinous process of each vertebra starting from the first vertebra at thoracic vertebrae towards the tail. The extracted points were matched to the predefined corresponding points of the spine in a rat atlas. Their correspondences field was used to perform skeletons registration by thin-plate-spline (TPS). Further, we extend 3D shape context algorithm for landmarks matching for rat limbs between the atlas and the newly extracted skeletons. The correspondences found were used to perform the rat limb skeletons registration by TPS.

Results Experiments were described using phantoms and actual rat skeletons. Mean square errors decrease was observed during registration process. Visually, the skeletons were successfully registered.

Conclusions The method can improve the robustness of rat skeleton registration, even in the case of large variation in some postures and this first step work can be extended to further improve rat organs registration guided by the correspondences found from the skeletons.