Iterative Gradient Vector Flow (IGVF) Method And Its Application In Medical Image Segmentation

Medical image segmentation is a major research field in image processing. Traditional segmentation models cannot satisfy the requirement of medical image segmentation due to its complexity. In recent years, a new segmentation method named deformable model was developed and was widely used in medical image segmentation. There are two types of deformable models: parametric deformable model and geometric deformable model. Recently Xu Chenyang of John Hopkins University developed the Gradient Vector Flow (GVF) parametric deformable model. GVF improves the performance of traditional deformable models by extending the external force to a much larger range over the image domain. Compared to traditional deformable models, GVF is much less sensitive to initial contour position and can converge into boundary concavity. But GVF still does not work very well in the case of long and thin boundary concavities.In this thesis, a new deformable model called Iterative Gradient Vector Flow (IGVF) is developed. IGVF uses GVF iteratively. In each iteration, the position of the resulting contour is adjusted and is then used as the initial contour for the next GVF iteration. Contour adjustment usually occurs near long and thin boundary concavities, where GVF would have trouble converging. Digital image processing techniques such as noise reduction and edge detection are also discussed in detail and used for preprocessing of medical images.Finally, IGVF and image preprocessing techniques are applied to segmentation of real medical images, including sections of human lower leg and human brain. The latter has a lot of complex boundaries with long and thin concavities. The results of the experiments show that, compared to GVF, IGVF can achieve better results with less number of iterations for images with long and thin boundary concavities.