Research On Segmentation And 3D Reconstruction Of Human Eye Tissues And OCT Images

The modeling and simulation of virtual organ is one of the frontier research areas. As an eyespot to observe objective things, eye is one of the most important organs of human. Modeling and simulation of eyes has very high value and application of scientific significance. With digital technology into the field of biomedicine, medical image processing is receiving increasing attention of researchers and exploring. Through the medical image processing can help doctors diagnose the disease, thereby guiding treatment and prognosis of disease, even cause analysis.In this paper, the main contents are concerned with the eyes of medical image processing, including: images segmentation of Human eyes tissue and radical OCT optic disc, registration of OCT optic disc images, middle interpolation, finally modeling.According to the characters of Human eye tissue images sequence and radical OCT optic disc images sequence from live human eye, we take proper algorithms to get the cornea from the tissue images and to get the optic disc from the fundus images automatic with high efficiency. And we study valuable registration for OCT images followed with middle interpolation and model methods. Finally, a demo system provides a graphics user interface to rotate, scale the 3D objects. It can extract images at any angle from the 3D model.The main contributions of this thesis are mainly lied in the following aspects:1,With the analysis of human eyes tissue images, a segmentation method based on BP neural network is proposed. Followed with the regional growth and holes filled, we can extract the cornea in different direction of images sequences.2,According to the characteristic of OCT optic disc images, an edge detection method is used to get the object effectively.3,Because images are radical scanned in the middle of optic disc hollow, a rigid registration method is applied with the help of curve fitting center-aligned.4,An interpolation arithmetic based on cubic spline function is designed to get intermediate images; To enable the reconstruction of the optic disc coherence, images are converted to cylindrical coordinates. The Bresenham algorithm is used to finish the interpolation among points; A 3D macula model is reconstructed finally. We can get optic disc images at arbitrary angle.