Medical Images Segmentation Using Contouring Techniques

Abstract

The underlying application domain for this work is medical imaging. Medical imaging allows scientists to observe anatomic structures inside the human body in a non-invasive way. The technological advances within medical imaging and medical image analysis have had a great impact on the medicine, both within clinical and research applications. It has expanded beyond simple visualization of anatomic structures and become a tool in surgical planning and simulation, intra-operative navigation, diagnosis, tracking the progress of a disease etc. As a tool in the medical research, medical imaging has for example become an irreplaceable aid in brain research applications such as cognitive psychology and morphological analysis of brain structures. As it is mentioned in the introductory part of this chapter, the problem of extracting information form images is the most important and challenging problem in the image analysis. Modern medical imaging devices such as CT and especially MRI, provide visually satisfactory images of the internal human organs. Problems arise when we want to utilize computers in order to automate the extraction of information from an image set.i have studied the characteristics of a brain image from an image processing point of view basically to detect tumor. The preprocessing of image is done to reduce the effect of speckles and preserve the tumor edges and thereby provide the foundation for a successful segmentation. The segmentation method we chose to implement were level set method- for it,s strength in segmenting volumes and the snake- which manages to conserve weak edges. These methods were implemented with Matlab from mathsworks. It was chosen for it,s wide range of medical image processing functions. As it is shown throughout this work, many active contour models give satisfactory results in some specific situations but fail to meet the various requirements e.g. they may be designed for a specific problem and can not be used on a variety of shapes, or the contour may be to rigid and therefore unable to capture complexity of anatomic structure. Another problem which arises in some models is a time consuming preprocessing of the images required to create a model capable of detecting a specific structure. Even though, these models give satisfactory results in some applications, they are not well suited for large scale automated processing of images in clinical applications. v In this work we observe representatives of different classes of active contour models, studied them theoretically, then choose two methods for implementation. What we have tried is to determine is how ―good‖ different active contour models are for medical image segmentation. The actual analysis in the medical imaging causes corrective action by taking proper medication whose decision is based on the contour analysis. Another original contribution of this work is the Fourier spectral algorithm for iterating the evolution equation, which has not been used in the literature. The results show that bath methods may successfully segment a tumor provided the parameters are set properly. Both methods required an initialization inside the tumor. Future work can be aimed at an algorithm to automatically estimate parameters, we believe automatic segmentation can be used to support the therapy and there by increase the quality of the treatment.