Search

Information such as changes in brain tissues, their area, number of lesions if they exit, size of lesions etc. is required for diagnosis of neurodegenerative diseases and also during treatment of these diseases. For this reason, skull, brain tissues and lesions are segmented manually from medical images; edges of these structures, number and size of lesions are determined subjectively. Visual examination and analysis of images is a time consuming and tedious task. Also, noise caused by imaging, and low contrast in the images make much more difficult the visual analysis and manual segmentation. This case, affects subjective evaluations, causes different decisions of different doctors on the same image, even different decisions of the same doctor on the same image at different times. Therefore, in this work, a computer aided approach that achieves automated segmentation of skull, brain tissues and lesions is proposed. In the proposed hybrid approach, magnetic resonance images have been used, skull and tissue segmentation has been performed by a probabilistic method based on Gaussian Mixture Model, while lesion segmentation has been performed by a deterministic method based on level set technique. By the developed software, lesion segmentation can be performed successfully (maximum symmetric surface distance is 5.76±3.24 mm). [ABSTRACT FROM AUTHOR]