Research Of OCT Fundus Image Segmentation Method Based On Improved Parametric Kernel Graph Cuts

Medicine is closely related to human health. In recent years, with the development of computer technology, Computer Aided Diagnosis has become an important auxiliary means of medical field. Computer Aided Diagnosis technology, which uses the ability of computer that can model accurately and process data quickly, combines with a variety of physiological and biochemical methods and medical image processing technology, can provide clinicians human organization lesion images and help doctors to diagnosis and treatment of disease. At the same time, the development of medical imaging technology has brought great convenience to clinicians. Optical Coherence Tomography is a new kind of tomographic imaging technique, which utilizes interference characteristics of coherent light, can imagine living miniature tissues and organs rapidly and noninvasively. Optical Coherence Tomography technology has been widely used in ophthalmology and gastroenterology in the clinical diagnosis and treatment due to its ability to achieve accurate and visual observation of human organs and tissues as well as no contact with the human body in the process of imaging.In the field of ophthalmology, the fundus images that could clearly display the information of retina can be an important basis for diagnosis of fundus disease. Fundus images that acquired by Optical Coherence Tomography imaging technology often carry some noise which might reduce the judgment accuracy of the disease. So to provide a satisfactory segmentation result of retina layers is in urgent need of clinical ophthalmology field.To solve this medical image segmentation problem, scholars put forward many solutions. Classic algorithm such as Canny edge detection algorithm, which through the calculation of the gradient of image matrix amplitude, with double threshold method to pick up the image grayscale changes in regional as the edge of the image information. Active contour model as a more mainstream segmentation algorithm, it through the internal force and deformation in advance the role of external forces to enter to divide the target contour model, so as to make the energy equation to minimum an then get the final segmentation result. In addition, normalized cut and level set algorithm can also complete image segmentation. However, the grayscale range of fundus image is limit, when dealing with Optical Coherence Tomography fundus images by the above algorithms, the results is not ideal.In order to overcome the shortcomings of existing methods, this paper proposes a solution combined with morphological erosion and parametric kernel graph cuts. This solution makes itself adapt to the characteristics of Optical Coherence Tomography fundus images by integrating and fine-tuning of the existing method. Morphological erosion that act as the image pretreatment process, can better remove the noise in the Optical Coherence Tomography images. As a widely used method of image segmentation, graph cuts algorithm add extra control points to divide each pixel of the image into two subsets named foreground and background. This article adopts the method of parametric kernel graph cut s to segment fundus image, parametric kernel graph cuts algorithm is an optimal segmentation strategy by fixing an alternately parameter of graph cuts method and optimize another. According to the characteristics of Optical Coherence Tomography, the peak value Signal-to-Noise Ratio is introduced to adjust the parameters of the energy function and process different quality of the fundus image automatically. The experimental results show that the proposed method can complete retinal fundus image segmentation of each layer.