Semi-automated Segmentation of Brain Tumor

  • Print

Segmentation of brain lesions from medical images is a difficult task to be mastered by the specialist. This is due to the presence of noise, partial volume effects and susceptibility artifacts in the images. These images also contain abnormalities in the distribution of the intensities of the white matter, gray matter and cerebrospinal fluid. All these problems can interfere with the results when manual segmentation is used. Manual segmentation uses local anatomical information based on the user experience; that implies the necessity of constant human intervention. Deformable model approaches (geometric and parametric) attempt to reduce these shortcomings by outlining the region of interest in a semi-automatic manner. These methods have been shown to be effective in the extraction of the lesion borders in brain MR images with reduced user intervention. However, due to the restrictions of the deformable models when dealing with regions without well defined edges, the proposal of this work is to apply the Mumford-Shah model via level set methods represented as geometrical deformable models, in order to segment multi-sequence magnetic resonance (MR) images of the brain composed of FLAIR (Fluid Attenuated Inversion Recovery), T1 and T2-weighted images. Results showed that segmentation using multi-sequence images provides superior results than using each sequence alone. As a part of this work, a software with a minimal human intervention has been developed to visualize and segment the brain lesions that appear as hyperintensities in MR images. As a consequence, medical doctors can exploit the segmentation results to follow up their patients by assessing the evolution or involution of the brain lesions.


Joomla SEF URLs by Artio