研究目的
To propose an automated algorithm to segment the inner limiting membrane (ILM) in spectral domain optical coherence tomography (SD-OCT) scans for the detection and diagnosis of glaucoma.
研究成果
The proposed automated algorithm successfully segments the ILM layer in SD-OCT images with an average error of mean Euclidean distance 7.029384 ± 3.72 compared to manual annotation. This segmentation is crucial for the diagnosis of glaucoma, which is the second leading cause of blindness worldwide. Future work includes the segmentation of the RPE layer and the calculation of CDR for glaucoma classification.
研究不足
The study is limited to the segmentation of the ILM layer and does not include the segmentation of other retinal layers or the calculation of cup to disc ratio (CDR) for glaucoma diagnosis. The algorithm's effectiveness is tested on a specific dataset, and its generalizability to other datasets is not discussed.
1：Experimental Design and Method Selection:
The study involves the development of an automated algorithm for ILM segmentation in SD-OCT images to diagnose glaucoma. The methodology includes image preprocessing, denoising, contrast enhancement, thresholding, ILM layer surface segmentation, and interpolation.
2：Sample Selection and Data Sources:
The dataset consists of 50 optic nerve head centered OCT images with a resolution of 456*951, obtained from the Armed Forces Institute of Ophthalmology (AFIO).
3：List of Experimental Equipment and Materials:
TOPCON’S 3D OCT-1000 camera was used for image acquisition. Illustrator CS6 was used for manual annotation of ILM layers.
4：Experimental Procedures and Operational Workflow:
The process includes cropping and rescaling images, quality assessment, denoising using bilateral and wiener filters, contrast enhancement, thresholding, objects removal, ILM layer surface segmentation, and interpolation.
5：Data Analysis Methods:
The performance of the algorithm was evaluated by comparing the automated ILM segmentation with manual annotations using Euclidean distance.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
