研究目的
To assess diabetic retinopathy by segmenting retinal layers in SDOCT images using axial gradient canny edge detection combined with a level set method and classifying abnormalities using SVM.
研究成果
The study successfully demonstrated that the RNFL complex is thinner in diabetic retinopathy patients compared to normal subjects. The proposed method of segmentation and classification using SVM with RBF kernel achieved an accuracy of 98.1%, assisting in the accurate diagnosis of diabetic retinopathy.
研究不足
The study is limited by the sample size and the specific conditions of the subjects. Future work could focus on automating the estimation of individual retinal layer thickness for earlier disease detection.
1：Experimental Design and Method Selection:
The study uses anisotropic diffusion filtering for speckle noise removal and axial gradient canny edge detection combined with a level set method for segmentation of RNFL, GCL, and IPL complex.
2：Sample Selection and Data Sources:
SDOCT retinal images of 75 subjects with uncontrolled diabetes and 30 subjects with controlled or normal diabetes from Aravind Eye Hospital, Puducherry, India.
3：List of Experimental Equipment and Materials:
SDOCT (Cirrus HD-OCT) machine with a scanning speed of 27,000 A-scans/s, MATLAB software, Intel (R) Core (TM) i7 CPU 860, 64-bit OS, and 8 GB RAM processor at
4：80 GHz. Experimental Procedures and Operational Workflow:
Images are filtered for noise, segmented using the proposed method, features are extracted, and classification is performed using SVM with RBF kernel.
5：Data Analysis Methods:
Statistical and run-length features are analyzed, and classification accuracy is evaluated.
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
