Abstract
Recent developments in medical imaging techniques have brought an entirely new research field. Medical images are frequently corrupted by inherent noise and artifacts that could make it difficult to extract accurate information, and hence compromising the quality of clinical examination. So accurate detection is one of the major problems for medical image segmentation. Snakes or Active contour method have gained wide attention in medical image segmentation for a long time. A Snake is an energy-minimizing spline that controlled by an external energy and influenced by image energy that pull it towards features such as lines and edges. One of the key difficulties with traditional active contour algorithms is a large capture range problem. The contribution of this paper is that to in-depth analysis of the existing different contour models and implementation of techniques with minor improvements that to solve the large capture range problem. The experiment results of this model attain high accuracy detection and outperform the classical snake model in terms of efficiency and robustness.
TopIntroduction
Medical image processing application has swiftly increased in recent years. Physicians and scientists make use of medical imaging techniques to visualize anatomic structures or mappings of physiological functions non-invasively with increased accuracy and precision. These tools influence areas including diagnosis, radiotherapy, surgical planning, and tracking of disease progress. The advent of various methods has provided physicians with powerful, non-invasive ways for studying the internal anatomic structures and physiological processes of the human body. The advances in imaging techniques also bring the benefit of providing better diagnosis and treatment options to many clinical applications. To facilitate visualization, manipulation, and especially quantitative analysis of medical images, methods are needed for the repeatable, accurate, and efficient localization and delineation of objects of interest from given medical images. Medical image processing and analysis covers a broad range of research topics, including image acquisition, formation, reconstruction, segmentation, filtering, enhancement, compression, and visualization.
With the rising availability of relatively economical computational resources like Computed Tomography (CT) Scan, Magnetic Resonance Imaging (MRI), Doppler ultrasound has all been necessary to the radiologists of imaging tools toward ever more consistent detection and diagnosis of disease. This is an actual time-consuming process that can take days to estimate a set of images from one patient. Furthermore, the boundary determination is made more challenging when the images are of poor class and the boundaries are difficult to see in the image. Medical images are often corrupted by inherent noise and artifacts that could make it difficult to extract accurate information, and hence compromising the quality of pathological examination. One of the challenging issues is the intensity in homogeneity, which can be observed as a relative variation of the intensity of the object to be detected. The boundary detection acting as a major role in the segmentation of medical applications such as abnormality detection, treatment progress monitoring and surgical planning.
Parametric Active Contour Model (PACM) was elaborated by Kass et al. (Kass, Witkin, & Terzopoulos, 1988) as a snake or active contour. After that, it has been effectively applied to a different category of problems in computer vision and medical image analysis such as motion tracking, edge and contours detection and segmentation. Snakes are also found to be practically useful in medical image segmentations. Examples include automated analysis of nerve cell images, lung segmentation (Ray, Acton, Altes, De Lange, & Brookeman, 2003) from magnetic resonance images, shape descriptions of fluid-filled regions from optical coherence tomography images of the retina (Ray, Acton, Altes, De Lange, & Brookeman, 2003), and detection of rolling leukocyte within in trivial microscopy. The main plus point of snake models, compared with conventional edge detection approaches are that they incorporate spatially and image information for the extraction of the smooth border of the region of interest. Initial points are estimation of the desired boundary is given, and the curve deforms to obtain the most favorable shape. Thus, isolated artifacts are ignored when they hamper with the smoothness of the curve.
The contributions of this chapter are listed here. An in-depth analysis of the existing different contour models is presented in this chapter. This chapter also includes the mathematical modeling of Active Contour Model with minor improvements of changes in the active contour model. Achieved improvements are tested on a large number of image sets.
Key Terms in this Chapter
Edge Detection: Edges are a partition between different textures of the objects in the class. Edge also can be differentiated as discontinuities in image intensity between two adjacent pixels.The edges of an image are always the important property that offer an indication of higher frequency. Detection of edges in an image may help for so many image processing applications like image segmentation, data compression, and also help for well matching, such as image reconstruction and so on.
Deformable Model: Deformable models are curves or surfaces defined within an image domain that provide an abstract model of an object class by modeling the variability separately in shape, texture or imaging conditions of the objects in the class. Deformable models denote the shape of objects as a stretchy 2D curve or a 3D surface that can be deformed to match a particular instance of that object class. It can move under the influence of internal force that are designed to keep the model smooth during the deformation and external force that defined to move the model toward an object boundary or other desired features within an image.
Parametric Active Contour Model: Parametric Active Contour Models (PACM) is a parameterized curve that is drawn outside or inside the object boundary that is to be segmented. This curve or contour is involved to or pushed away from certain features of the object in the image. This can be compared with an elastic band that is stretched out and tries to contract back into its original shape against a force. The flexible represents the contour, and the energy is derived from the image.
Image Segmentation: Segmentation is the method that is dividing an image into separate regions containing each pixel with similar attributes. To be meaningful and useful for image analysis and interpretation, the regions should strongly relate to depicted objects or features of interest.
Greedy Approach: A greedy algorithm is a mathematical procedure that looks for simple, easy-to-implement solutions to complex, multi-step problems by deciding which next step will provide the most obvious benefit.
Medical Image Processing: Medical image processing is the technique and process of creating visual representations of the interior of a body for the scientific analysis and medical intervention. Medical imaging seeks to reveal internal structures hidden by the skin and bones, as well as to diagnose and treat disease.