Research On Local Segmentation Of Basal Ganglia Region In Brain MR Images

To accurately segment some cerebral nucleus of basal ganglia region in brain MR images, a local predictive active contour model and a shortest path based interactive segmentation method are proposed in this paper.As local information near the contour is usually used to control the contour evolution in local active contour model, considering the inhomogeneity and the gradual change characteristic of MR images in the interior area, local information of larger interior area is adopted to predict and estimate the feature of the area which will be evolved. Then comparing the errors between the predicted feature and the feature calculated from exterior area, the weak edges of MR images with intensity inhomogeneity are effective to be identified using the local predictive active contour model proposed in the paper. The experimental results indicate that the proposed model can effectively deal with weak edges for white matter segmentation with intensity inhomogeneity. For segmenting putamen, the segmentation accuracy of the local predictive method is acceptable; but its stability needs to be further improved.To accurately segment putamen in brain MR images, a shortest path based interactive segmentation method is proposed. The cost function uses weight coefficients in sub regions and orientations, and the shortest path search region is restrained, so that the method can realize fast segment for putamen. Great deals of practical experiments show that the proposed shortest path segmentation method not only realizes the fast segmentation for putamen, but also guarantees the accuracy, stability and practicability of the segmentation.