Study On SS-OCT Retinal Imaging System And Its Image Segmentation Algorithm

The retina,located in the innermost layer of the entire ocular wall,is the only internal physiological structure of the human body that can be non-invasively observed from the outside.The early clinical symptoms of a variety of diseases in the human body are reflected in the retina,so various diseases can be found and diagnosed in the early stage by retinal examination.Swept-Source Optical Coherence Tomography(SSOCT)has been widely used in the clinical diagnosis and treatment of retina due to its non-contact,non-invasive,and radiation-free safety features,as well as high resolution and high detection sensitivity imaging features.In this thesis,the SS-OCT retinal imaging technology is deeply studied.firstly,based on the principles of SS-OCT imaging and the characteristics of retinal imaging technology,a set of SS-OCT retinal imaging systems is designed and constructed,which realizes the retinal tomography of rabbit eyes.Secondly,an algorithm suitable for Cystoid Macular Edema(CME)lesion segmentation in OCT images is proposed,which can quickly extract the contour of CME and provide more reliable auxiliary diagnostic information for clinicians.The specific research contents are as follows:1.The SS-OCT retinal imaging system is designed,which includes a frequency scanning light source module,an OCT imaging light path module,a galvanometer control module,and a data acquisition and processing module.According to the requirements of retinal imaging,the structure of the front objective lens group and the control method of scanning galvanometer is designed.The upper computer software was written with C# and realized many image processing functions,such as patient filing,imaging mode selection,hardware parameter setting,real-time imaging monitoring.2.The existing clinical instruments cannot automatically identify CME lesions,a full-automatic segmentation algorithm based on an omnidirectional wave operator was proposed to solve the problem.The algorithm is based on the wave algorithm applied in the field of retinal layer segmentation,and the directional adjustment function is designed to endow it with multi-directional attributes.By using the gray gradient characteristics of the boundary region in OCT images,the cyst region can be quickly and accurately extracted through the steps of constructing the working environment,extracting the ROI region,extracting contour,and integrating contour.3.A SS-OCT retinal imaging system was constructed,and the system-related parameters were measured,followed by animal imaging experiments and CME segmentation experiments.The experimental results show that the system can achieve tomographic structure imaging of fundus of living animals.The algorithm proposed can accurately extract the contour of cystoid macular edema.Compared with the commonly used DRLSE segmentation algorithm,the anti-missing detection ability is increased by36%,the F1-Score is increased by 25%,and the operation speed is increased by 15 times.Compared with the commonly used Snake segmentation algorithm,the antimissing detection ability is increased by 12%,the F1-Score is increased by 6%,and the operation speed is increased by 40 times,which proves that this algorithm can provide a feasible auxiliary tool for the clinical diagnosis and treatment of cystoid macular edema.