Vascular segmentation in magnetic resonance angiography: A modified region growing approach

A modified region growing algorithm is proposed to extract cerebral vessels using a magnetic resonance angiography (MRA) database. To improve the performance of the image segmentation method, as a pre-processing step, image enhancement methods are applied by the gamma correction technique and spatial operations. This step improves the detection of gray-level discontinuities in MRA images. The traditional region growing method is modified by extending the neighborhood as 24 pixels and by defining a filling protocol to label vascular structure. The performance of the proposed algorithm is compared with that of the traditional region growing method and four other segmentation methods. The minimum and maximum errors of the modified region growing algorithm is calculated as zero and 1.11%, respectively while the traditional region growing method has 0.2% and 7.81%.