Automated Segmentation of Brain Structure from MRI
L. Amini, Hamid Soltanian‐Zadeh, Carrie L. Lucas · 2003
Synopsis We developed a new method for segmentation of specific brain structures from MRI. The method is based on a dynamic contour model, which deforms under external and internal forces. The desired image features are utilized to define external force. A new algorithm integrating thresholding, fuzzy clustering, edge filters, and morphological operations overcomes limitations of low contrast tissues and unclear edges. A new method automatically generates the initial contour for the model, to delineate dependency of results on the operator. Application of the methods to thalamus illustrates its steps and similarity of the results to manual segmentation by a radiologist.