Thalamus is one of the important brain Structures, which affects many parts of the brain cortex. Segmentation of this structure and extraction of its features are expected to aid with the evaluation of the relative illnesses and treatments noninvasively. We have developed a fully automatic method for segmentation of thalamus. We process neuroanatomic and image data by a multiple step algorithm. This processing implements perceptron algorithm, radial basis function (RBF) map, fuzzy cmeans (FCM) method and landmarkguided boundary detection (LGBD) algorithm. We fused neuroanatomical and image information to obtain some thalamic and nonthalamic points, then we applied RBF map and perceptron algorithm to get the coarse location of thalamus which we call our region of interest (ROI). Thalamus is relatively hard to segment for its low signaltonoise, low contrastonoise ratio and discontinuous edge. The first two problems has been solved by improving the ROI. Appling clustering algorithm like FCM on the ROI, increases the accuracy of segmentation compared to its applying on the whole image, therefore we can obtain some parts of the border. The new LGBD algorithm has been tacked the problem of edge discontinuity. It completes the border by the results of the pervious steps of the main algorithm. The method is tested on real clinical data and the results show very good performance.
Download TeX format
