Study Of Automatic Algorithm For Registration And Background Removal In Slice Images Of A Mouse

With the development of life science and information technology, the digitalization of living organisms and functions has become the focus of life science research in recent years. As the most frequently used animal in medical experiments, the development of mouse model will bring a breakthrough in medical and human health research. Prior to the 3D reconstruction, it is crucial to implement the accurate registration and background removal in slice images of a mouse, especially when the significant defects are generated in the slice images during the production process. Moreover, since the mass data processing is always involved in the 3D reconstruction from slice images, the background removal is also an important issue from the viewpoint of reducing data redundancy effectively.In this thesis, the slice image dataset of a mouse was taken as the research object , which were collected by Third Military Medical University in 2007. Based on the analysis of the problem of this data sets and the characteristic of image background, the author proposed an automatic registration and background removal method. (1) Image registration: an external registration algorithm based on calibration rods modification was put forward. Firstly, an external calibration rod was extracted by means of edge detection, filtering etc., and the centre points'coordinate of the calibration rod were calculated by referring to the definition of the image matrix. Then the centre point of the external calibration rod at the lower left corner was used as the reference point to eliminate the sloshing of the slice image sequence in the horizontal direction. Finally, the least square method was adopted to calculate the scaling factor for each slice image, based on the distance between the reference point at the lower left corner and the other one at the upper right corner of the image, so that the problem of gradual image reduction of the target region was solved. (2) Background removal: In the light of the image background characteristics and the fact of gradual variation in the slice image sequency, a method based on the template matching theory to remove the background was put forward in this study. It was implemented by six steps, including creation of a template, formation of a ring-shaped area, extraction of edge contour, deburring based on distance and slope, elimination of discontinuity by means of the Lagrange interpolation, and the background removal by comparing the refilled contour and the original image. In the research, the contours could not be extracted automatically by computer, because the contour of mice'head is changing so quickly and the contour of limbs is not connected domain with body. The mice slice data sets were segmented and the contours were modified manually, in this way the integrity of mice slice data sets was guaranteed.In addition, the registration results were evaluated in two methods. The first one involved the comparison of the slice image reconstruction effects before and after the registration, while the second one involved the ues of three statistical parameters, which were root mean square error, peak signal to noise ratio and cross entropy. The results show that the proposed method offers a solution to the problems of sloshing of the slice images in the horizontal direction and image reduction of the target region, and allows the accurate registration. As for the results of the background removal, they were assessed by the comparison between the original images and the processed images, as well as the observation of the reconstructed 3D model. In conclusion, the results indicate that the automatic algorithm for background removal performs well, and the obtained contours of the mouse are smooth. Consequently, the workload in relationto the manual background removal would be remarkably reduced and the three-dimensional reconstruction would be evidently sped up.