3D medical image segmentation technique

Despite continuing advances in mathematical models for automatic segmentation, many medical practitioners still rely on 2D manual delineation, due to the lack of intuitive automatic tools in 3D. In this paper, an efficient 3D medical image segmentation technique is proposed to provide 3D representation of the segmented regions. It uses graph cut and contour filling algorithms. It uses the normalised cut method with the eigenvector of the second smallest eigenvalue to solve the image segmentation problem, and the contour filling algorithm to ensure that the segmented region is free of gap and hole artefacts. The experimental results reveal that the proposed technique can provide a 3D representation of the region of interest successfully. The segmentations produced by this method are more realistic than the previously proposed segmentation techniques besides its effectiveness in reducing the amount of gaps and holes.