| Glaucoma is the leading cause of irreversible blindness in the world. Primary angle closure glaucoma is the predominant glaucoma subtype in Asians. According to the pathogenesis of primary angle-closure glaucoma, assessment of anterior chamber angle (ACA) is an important component in glaucoma diagnosis. Optical coherence tomography (OCT) is a non-contact technique and can be performed easily with minimal expertise. These advantages make OCT suitable for ACA imaging. The angle opening distance (AOD) measurement is a common and effective method for ACA assessment. Traditional AOD measurements from OCT are based on manual or semi-automatic methods, which are time consuming and also operator-dependent. Besides, the state-of-the-art current ACA OCT modalities only obtain a2-dimentional (2D) cross section image at one location angle in a single scan. It can not provide a full range of angle in a quadrant as gonioscopy does. Therefore, how to measure AOD automatically, and get more cross section images at different location angles in a single scan are significant reserch topics.In order to overcome the limitations of the traditional AOD measurement in OCT images, as single scan modality and inefficient measurement method in2D ACA images, an ACA segmentation algorithm based on area enhancement in the OCT image is proposed. Based on this algorithm, a new automatic AOD-SL measurement method for2D ACA OCT image is developed. The agreement between automatic and manual measurement results is quite good. Furthermore, a3D scan pattern is proposed through a multi-angle imaging. A synthetic B-scan and AOD-SL in each scaning angle is obtained by image processing, which can give more angle information for glaucoma diagnosis in clinic. This research includs several key techniques used in computer aided glaucoma assessment, such as ACA segmentation. AOD-SL automatic measurement and3D angle measurement. It provides the reference of theoretical basis and relevant techniques for popularizing computer aided primary angle closure glaucoma measurement and evaluation. The main researches are descriped as follows.To evaluate the raw and the enhanced OCT image quality, a objective assessment method is proposed. Three parameters, including the noise variance, spread of edges and contrast-to-noise ratio are used. At the same time, a subjective grading standards was made, according to the edge sharpness, continuity and visibility of retinal layer structure in the OCT image. The assessment reslut shows the proposed objective assessment method is an effective way for image quality evaluation. It can be a new solution for the valid diagnostic image selection in clinic.Considering the poor quaily of raw ACA OCT images with speckle noise, shadow, artifacts and low contrast, a new ACA segmentation algorithm with area enhancement is proposed. Firstly, the image is filtered for the speckle removal. Secondly, the angle region in the image is enhanced and recovered by using the morphological grayscale reconstruction and gamma correction. Finally, the segmentation is done with artifacts removed. The result shows, this algorithm is robust and accurate for ACA image with different image quality. This new segmentation method provids a practical way for precise computer aided measurement of AOD.An advanced automatic measurement method of AOD based on Schwalbe’s line (AOD-SL) is presented. The image is segmented firstly. Secondly, boundaries lining the anterior corneal, corneal endothelium and anterior iris surfaces at the iridocorneal angle are detected, and the image is dewarped. Thirdly, posterior corneal Descemet’s membrane is obtained in the gradient image based on corneal endothelium. and Schwalbe’s line is detected according to the distance between the corneal endothelium and Descemet’s membrane. Finally, the AOD-SL can be obtained based on the Schwalbe’s line and anterior iris surface. Compared with manual detection of Schwalbe’s line, the automatic detecton accuracy rate is90.38%. The agreement between two dection methods is high, and their Pearson correlation coefficient is0.99. Furthermore, the automatic detection results is positively correlated with gonioscopy grades. Thus, this proposed method is robust and accurate. It is more effective and objective than the traditional methods.To overcome the limitation of existing2D ACA OCT scan solutions, a completely novel3-dimensional (3D) scan pattern of Fourier domain OCT imaging is designed, and a synthetic B-scan is redrawn. Also an3D AOD-SL measurment method is proposed based on the synthetic B-scan. The method first segments25frames in a3D image set by using algorithm proposed in chapter3. Boundaries lining the anterior corneal, corneal endothelium and anterior iris surfaces at the iridocorneal angle are then detected. A straightforward way to speed up the detection is to apply initial boundary detection in a reduced image. And then final boundary can be obtained based on the initial results in a gradient image. In this step, the ’dewarping’ is made for refraction correction. After the positions of Schwalbe’s line in the three key frames are located according to user input, the Schwalbe’s lines in the other22raw frames can be calculated by mapping algorithm. With that a synthetic B-scan can be obtained due to the known SL position, and surface of anterior iris. Finally, the AOD-SL is calculated from the synthetic B-scan. The result shows, this3D measurement method is accurate at boundary dection and Schwalbe"s line located in raw frames.’The agreement between the3D measurement and the2D manual measurement is high, and their Pearson correlation coefficient is0.91. Furthermore, the automatic detection results is positively correlated with gonioscopy grades. Thus, this proposed3D method is robust and accurate. It can provide more angle information for clinic. |
PDF Full Text Request