Research Of Glaucoma Retinal OCT Images Analysis

The retina is composed of delicate and vulnerable tissues,and is the target organ of many diseases.Glaucoma is recognized as a progressive optic neuropathy and is the main leading cause of blindness both in China and worldwide.So far there is no clinical method to reverse the optic nerve damage.Therefore,early diagnosis and treatment are essential in reducing the risk of visual loss caused by glaucoma.As a real-time,in-vivo,noninvasive and high resolution imaging technique,optical coherence tomography(OCT)has been popularly used in monitoring retinal structure changes,and in treatment guidance and evaluation.To deal with vast amounts of images,efficient methods to analyze these retinal OCT images are in great need.As a result,retinal OCT image processing and analysis has been a hot but challenging topic in the field of medical image analysis in recent years.This dissertation mainly studies the methods of retinal OCT images segmentation,retinal thickness analysis and retinal optical intensity analysis based on both normal and glaucoma subjects.The main contributions of this dissertation are summarized as follows:1.A statistical model and graph search based algorithm is proposed that can automatically detect the margin of optic disc and optic cup in ONH centered 3D OCT scans.The complex structure,abundant vessels and nerves make segmentation of optic disc and cup a difficult problem.In the proposed algorithm,the optic disc is first localized based on statistical models.Then the disc and cup boundaries are optimized by the graph search method.The algorithm efficiently utilizes both the shape and texture information to achieve accurate segmentation of the optic disc and cup in ONH centered 3D OCT scans.The experimental results showed good performance.2.An algorithm is proposed that can automatically segment the intraretinal layers in optic nerve head(ONH)centered 3D OCT scans.Aiming at the problem of segmenting the multi-layered structure with a "hole",we proposed the GS-AAM algorithm,which combines the active appearance model(AAM)and graph search(GS)method.The AAM can fully capture the 3D structural information around the ONH.Based on the constraints obtained by AAM,the GS further finds the optimal result with a given cost function.Accurate segmentation can be obtained finally.3.The performance of our graph search based algorithm and Topcon’s algorithm are compared and analyzed.We compared these two algorithms by analyzing the retinal thickness on normal subjects and primary open-angle glaucoma subjects.The result showed that our graph-based algorithm exhibited a high degree of agreement with the Topcon algorithm with respect to thickness measurements for both normal and glaucoma subjects.Moreover,our algorithm can segment the retina into more layers than the Topcon algorithm does,and more accurately.4.The characteristics and determinants of retinal optical intensity distrbution in normal subjects and the significance of the retinal nerve fiber layer(RNFL)optic intensity in glaucoma diagnosis are studied,based on 3D OCT images.It was found that,the optical intensity was the highest in the RPE,photoreceptor layer(PR),RNFL,but the lowest in the outer nuclear layer(ONL).The optical intensity was the lowest in the central area of RNFL,ganglion cells layer(GCL),inner plexiform layer(IPL),inner nuclear layer(INL),outer plexiform layer(OPL)and photoreceptor layer(PR),but the highest for central RPE.In most layers,the optical intensity was positively correlated with image quality,and negatively correlated with age.There was no statistical correlation between retinal optical intensity and sex,height,axial length,etc.It was also found that,the optical intensity of RNFL was negatively correlated to the severity of glaucoma,and significantly associated with visual field defect.Our research can provide the basis to establish normal optical intensity atlas.The study of RNFL optical intensity gives a promising parameter for glaucoma diagonsis.