Research On Segmentation Algorithm Of Mouse Images Based On Deformable Mouse Atlas

Small animal image analysis plays an important role in preclinical cancer and new drug research,and mice are the most commonly used in small animal experiments.The most common way is to perform CT imaging on mice.Therefore,the processing of mouse CT images has become the focus of small animal image analysis.The use of digital anatomy atlas is an important way to provide anatomical structural references to target images through atlas registration.This research focuses on the micro computed tomography(micro-CT)of mice,and register the deformable mouse atlas to the target images to classify the organ regions of the target mouse.Building the deformable mouse atlas is the first part of this article through C++.The atlas can change posture,spine length,and weight.The individual morphological differences between organs are represented by a statistical shape model(SSM)during deformation.The positional relationship between the high contrast and low contrast organs can be represented by the conditional Gaussian model(CGM).The C++ version atlas has a faster deformation speed and is easy to publish.Based on the deformable mouse atlas,this paper proposes an atlas registration algorithm via high contrast organ segmentation.This method registers the high contrast organs of the atlas to the target images based on the segmentation results of these organs,it then map the low contrast organs to the target to achieve the registration of whole-body atlas.However,this algorithm is highly dependent on the segmentation of high contrast organs and cannot be used for images with too low resolution.In order to overcome the limitations of this algorithm,another improved algorithm is proposed,which is an atlas registration algorithm that does not require high contrast organ segmentation.The high contrast organs of the atlas can be filled with CT values,and get the filled grayscale image.The filled image is directly registered to the micro-CT image and get the final registration.The improved algorithm not only removes dependence on high contrast organ segmentation but also improves registration accuracy and robustness.With the increase of mouse CT images,various morphological postures appeared in the image.In order to analyze these data,this paper proposes a new atlas registration based on the anatomical landmarks.This method adjusts the atlas morphological posture by manually specified landmarks,and perform the subsequent registration.According to visual assessment,the new algorithm can get satisfactory registration results.This method was quantitatively compared with the registration method using the well-known Digimouse atlas using three metrics,i.e.,Dice coefficient,recovery coefficient of organ volume,and surface distance.The algorithm has better performance,higher registration accuracy,and better robustness.In this paper,the deformable mouse atlas was used for registration with the micro-CT images of mice.It innovatively achieved the registration of low-resolution images and solved the problem of changes in body posture and Individual organ morphological differences based on the atlas deformability while ensuring the registration accuracy and robustness.