Research On Graph Min-cut Theory And Its Applications In Medical Image Segmentations

Retinal and degeneration optic nerve diseases(such as glaucoma, diabetic retinopathy, age-related Macular disease, etc) are caused by a variety of retinal and neural tissue lesions.The kinds of degeneration diseases in the world have high incidence rate, and diagnosis in early stage is key point. At present, most of the retina (optic nerve) images are captured by the fundus color stereo cameras, which are of poor quality, very low contrast, obscure due to blood vessels, and distinct inter-differences of individuals. To segment and measure the optic nerve head medical images for the computer aided diagnostics of glaucoma diseases, a graph min-cut method is proposed based on PCLS M-S model.Firstly, we get the cup and the disk’s morphological edge via the segmentation of the optic nerve head image by using graph min-cut method for PCLS Mumford-Shah Model. Next, the optic cup and disk characteristic rectangles and edge points are extracted based on the color reconstruct image of an optic nerve head by incorporating the prior knowledge of the optic cup and disk shapes. And then, we use the technique of smoothing spline curve fitting to reconstruct the edge of cup and disk by vascular occlusion. Finally, we reconstruct the optic nerve head’s cup and disk image by using the reconstructed edge of cup and disk, and easily extract the pathological features parameter such as cup disk ratio based on the reconstructed image.The tests with the color optic nerve head images of different glaucoma patients showed that this method is able to handle this kind of images, which are of poor quality, very low contrast, obscure due to blood vessels, and distinct inter-differences of individuals, and to effectively segment the optic nerve head of glaucoma patients. We find that the proposed method is more efficient and more accurate than the piecewise constant level set method (PCLSM) for multiphase Mumford-Shah model.The graph min-cut method for PCLS M-S model segments the image by only using a single level set, and solves the resulting minimization problem by the graph min-cut combinatorial optimization technique. There is also no need of manual intervention.