Real Time Segmentation Technology Of Fundus Retina In Optical Coherence Tomography

The analysis of retinal images is the key process of research and diagnosis of ophthalmic diseases.The morphological changes of each layer of retina can reflect the health status of eyes,which is an important basis for screening high-risk groups and timely treatment of patients.Optical coherence tomography（OCT）is one of the most effective imaging technologies in ophthalmology.It has the characteristics of high resolution and non-invasive.After using OCT system to collect retinal images,it is usually necessary to rely on doctors for artificial retinal image analysis,which is a very energy-consuming and inefficient process.Therefore,it is of great significance to improve the efficiency of diagnosis while ensuring the accuracy.To solve the above problems,the following work is carried out in this thesis:（1）A 1060 nm center wavelength swept source OCT（SS-OCT）system is designed and built.Based on this SS-OCT system,the images of retinal macular region were successfully collected,in which the signal-to-noise ratio of the system was 106.7 d B;The image resolution is 1168 × 852;the imaging depth is 4 mm;The average irradiation power of the sample arm is 2.5 m W,which meets the national safety standard.The accuracy of SS-OCT system is verified by using the method of optical surface spacing measurement.The final measurement error is 0.013 ～ 0.51 μm.It ensures that the division of retinal layer is not affected by the system resolution.（2）The shortest path algorithm based on graph theory is optimized to improve the computational efficiency.Using the shortest path algorithm,the collected retinal images and diseased retinal images are processed in layers by means of edge recognition and search area restriction.The seven main retinal layers are successfully divided,which improves the diagnosis efficiency.Compared with manual stratification,the average deviation of automatic stratification is 2.81 ～ 5.63 μm.Within the acceptable range of clinical diagnosis.Combining the acquisition terminal of SS-OCT system with automatic layering software,retinal image acquisition and real-time layering are realized.（3）The retinal fundus images with different resolutions and different diseases are processed successfully,and the applicability of the layering algorithm is verified.Finally,based on the layered two-dimensional retinal slice data,the three-dimensional reconstruction of OCT fundus retinal slice image is realized by using light projection algorithm and Open Graphics Library.The three-dimensional model of each layer of retina is successfully separated.After rendering,the shape of retinal layer is more conducive to recognition and observation,which further improves the diagnostic efficiency.