Segmentation of Medical Images using Image Registration?

Medical image segmentation is one of the most essential task in many medical image applications, as well as one of the most complex tasks. Medical image segmentation aims at partitioning a medical image into its constituent regions or objects, and isolating multiple anatomical parts of interest in the image. The precision of segmentation often determines the final success or failure of the whole application. For example, when doctors want to reconstruct a 3D volumetric model of the heart, they need to segment the regions of heart in a series of 2D images. If segmentation is done wrongly, the reconstruction will be erroneous. Therefore, considerable care should be taken to improve the reliability and accuracy of segmentation in medical image analyzing and processing. If the region of interest in image have homogeneous visual feature then the segmentation is very easy. However, in more general medical applications, images are much more complex, and difficulties exist inevitably in segmenting these images. The difficulties of medical image segmentation are mainly based on the nature of imaging technology, dealing with low contrast image with noise, image properties, overlapping parts of an image. Due to these difficulties, intelligent algorithms are needed to segment multiple anatomical parts of medical images. One promising approach is registration-based segmentation. A model of the anatomical parts of interest is constructed. The model is registered to the image of a patient. When registration is correctly performed, segmentation of the various anatomical parts is done. By representing prior knowledge in the model, registration-based segmentation can handle complex segmentation problems and produce accurate and complete results automatically.