研究目的
To develop an integrated system design platform that efficiently detects the abnormal thin vessels (neovascularisation) and the lesions for gradation of DR under CS framework.
研究成果
The proposed method efficiently detects abnormal thin vessels and lesions for gradation of DR under CS framework, offering improved performance over existing works. It achieves high sensitivity and specificity at 80% measurement space, making it suitable for practical applications like MRI on the eye and tele-diagnosis.
研究不足
The study's performance is dependent on the measurement space, with lower spaces leading to increased noise and false detections. The method's complexity and computational time may also be limitations.
1：Experimental Design and Method Selection:
The study uses compressed sensing for image reconstruction, fuzzy entropy maximisation for vessel extraction and classification, and mutual information maximisation for neovascularisation and lesion detection.
2：Sample Selection and Data Sources:
Retinal images from DRIVE, STARE, DIARETDB1, and MESSIDOR databases are used.
3：List of Experimental Equipment and Materials:
The study involves image processing techniques and algorithms without specifying hardware.
4：Experimental Procedures and Operational Workflow:
The process includes image reconstruction, monochromatic image selection, pre-processing, optic disc removal, matched filtering with Gaussian kernel and scale estimation, LoG filtering, neovascularisation detection, lesion detection, and post-processing.
5：Data Analysis Methods:
Performance is evaluated using sensitivity, specificity, accuracy, and ROC curves.
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
