Local optimal threshold segmentation and reconstruction of cerebrovascular MRA images

Otsu algorithm is a common method in adaptive optimal threshold selection for image segmentation. It is used to calculate the optimal threshold of image segmentation by the maximal between-class variance automatically. However, the Otsu method highly depends on the distribution of the valley-peak in the image histogram, which may lose details in the segmentation of cerebrovascular in magnetic resonance angiography (MRA) image. According to the anatomic structure of cerebral vascular and the characteristics of the MRA images, we use a method that segments the cerebral MRA data-sets by combining with contrast-enhanced and local adaptive threshold, and then reconstruct the segmented image sequence using the Visualization Toolkit (VTK). The experimental results demonstrate that the applied method can compensate the weakness of the Otsu method on cerebrovascular segmentation of MRA images.