Research On The Algorithm Of Corpus Callosum Image Stitching And Alignment In Autistic Mice

Synaptic scale imaging and three-dimensional reconstruction of neural connections in nanoscale can directly display the physical structure of neural circuits,establish optimized disease model,and help to further explore the relationship between neural circuit structure and brain activity.In recent years,many researches have been carried out on the ultrastructure of exploring organisms at home and abroad,and different algorithms have been used to reconstruct different images in nano scale.With the development of ultra-high resolution technology,electron microscope(EM)has high resolution,clear cell structure and good imaging quality.For the anatomical reconstruction of large biological specimens,serial section microscopic imaging is a classic technology,which can achieve multi-resolution imaging.However,this method may introduce mechanical deformation,staining artifacts and other non ideal factors in the imaging process.Therefore,we need to use image stitching and alignment technology to maximize the preservation of the original tissue morphology.According to the characteristics of scanning electron microscopy(SEM)imaging of serial sections of corpus callosum in autistic mice,the image preprocessing,stitching and alignment algorithms were studied.Based on the information entropy and gradient information,the fuzzy image is filtered,and the histogram equalization method is used to adjust the image contrast to complete the preprocessing.Secondly,SIFT(scale invariant feature transform)algorithm is combined with RANSAC(random sample consensus)algorithm to extract relatively accurate feature points for feature matching,and then the weighted fusion method is used to realize the two-dimensional stitching of the corpus callosum electron microscope image of autistic mice.Then,by combining the elastic and rigid alignment algorithm,the elastic alignment based on the spring molecular system is constructed to remove the elastic deformation between images,and the least square method is used to estimate the rigid transformation between images to remove the rigid deformation,so as to realize the three-dimensional alignment of serial slice images.Finally,the results of alignment were analyzed qualitatively and quantitatively by using the relevant alignment quality evaluation indexes,and the reconstruction of corpus callosum in autistic mice under electron microscope was realized.The experimental results show that,compared with the subjectivity brought by the naked eye detection of fuzzy images,the proposed algorithm can objectively detect the fuzzy images,with high accuracy and improved time.It only takes 3 minutes to determine 600 slices.Based on the combination of SIFT and RANSAC algorithm,the feature points can be extracted relatively accurately for stitching,and then the three-dimensional image of part of corpus callosum of autistic mice is reconstructed successfully by combining elastic-rigid alignment algorithm.The similarity measure function RMSE(root mean square error)and SSIM(structural similarity)are selected as the quantitative indexes to judge the alignment accuracy.Compared with the rigid alignment,RMSE and SSIM are improved by 7.32%and 6.6%,respectively;compared with the elastic alignment,RMSE and SSIM are improved by 2.47%and 1.5%,respectively.The acquisition of three-dimensional images of part of corpus callosum in autistic mice lays a foundation for the segmentation and measurement of the tissue structure of corpus callosum,and the subsequent study of the relationship between corpus callosum lesions and the pathogenesis of autism from the perspective of ultra pathology.