Research On Automatic Segmentation Algorithm For Cerebral Cortex Of Digital Human Brain Slice

Digitized Virtual Human Project is a long-term scientific planning, which isrepresentative in cutting-edge technology and its development will greatly promotemedical science from subjective quantitative description to the deep development ofdigital instructions. Building a digital human brain model is critical of digitized virtualhuman, in which the first step is regional segmentation of the primitive brain slices.Now hand-sketched or semi-automatic segmentation is mainly used to extract theorgans and tissues of the human slice images at home and abroad, apparently it’s verytime-consuming and labor-intensive for processing massive human slice dataset and thequality of the segmentation results is greatly decided by clinical anatomy experience.Therefore, the urgent need for body color slice datasets to develop a segmentationalgorithm with higher degree of automation.In this paper, review and analysis of medical image at home and abroad isconducted, especially applied to the brain image segmentation algorithm. And forhuman brain slices in dataset of first Chinese visible human female, two segmentationalgorithms to accurately extract cerebral cortex from brain tissue are proposed:①Based on double edge detection and region growing segmentation algorithm. Asa preliminary exploration of the cerebral cortex segmentation algorithm, the algorithmfirstly makes joint use of RGB space edge detection and canny detection operator to gettrue continuous brain tissue contours, which can improve the traditional operator thatcan only get the wide edge. And then grayscale image of the brain tissue is obtained bythe region filling, corrosion and the mapping between images. In the end, using regiongrowing method which is based on the seed point obtained automatically segments198cerebral cortex results.②The joint with HSI space region growing and FCM clustering segmentationalgorithm for block matrix. The algorithm apply segmentation approach step by step,based on H channel information in the HSI color model which is able to describe thecharacteristics of the solid color attribute, H channel grayscale images is extracted afterHSI color model transformation firstly. Then the ideal primary segmentation result isobtained by using the region growing and image maps based on the above. Finally, theblock matrix cluster center of fuzzy C-means clustering is determined when image issplit into block matrixs, the algorithm achieves continuous segmentation of the cerebral cortex for400human brain slices.Further validate the effectiveness of the segmentation algorithm, in this paper thequalitative description and similarity coefficient DSI quantitative calculation of cerebralcortex segmentation results is conducted, in which the expert manual is referencestandards, three-dimensional reconstruction for all the results of cerebral cortexsegmentation sequence is obtained successively. Experimental results show that both oftwo segmentation algorithms provided in the article can achieve effective automaticsegmentation of the slice datasets. In contrast, the details reflect for cortical sulci andgyri is much richer by the second segmentation algorithm, and the algorithm has sostrong flexibility that it can obtain cerebral cortex of all brain slice images, whichprovide the appropriate conditions for further study of the human brain reconstructionmodel.