研究目的
To locate the optic disc and detect the optic disc boundary accurately in retinal images.
研究成果
The proposed method is effective for fast, automatic, and accurate OD localization and segmentation, achieving high accuracy and robustness across various retinal images. It contributes to improved evaluation of the OD and supports the development of automatic screening systems for early eye diseases. Future work could integrate ellipse shape constraints into the level set model for further reliability.
研究不足
The method may fail in cases with large, bright regions near the OD or severe pathological changes that make the OD dark and blurry. Computational efficiency depends on grid sampling size, and there is a trade-off between accuracy and processing time.
1：Experimental Design and Method Selection:
A multi-stage strategy is used, involving OD candidate identification based on brightness and vessel convergence, a line operator filter for accurate localization, iterative thresholding and ellipse fitting for initial contour, and a region-based active contour model with level set formulation for boundary estimation.
2：Sample Selection and Data Sources:
The MESSIDOR database with 1200 fundus color images of the posterior pole, captured using a Topcon TRC NW6 non-mydriatic retinograph with a 45° FOV, at sizes of 1440*960, 2240*1488, or 2304*1536 pixels.
3：List of Experimental Equipment and Materials:
Topcon TRC NW6 non-mydriatic retinograph for image capture; computational methods implemented in software (specifics not provided).
4：Experimental Procedures and Operational Workflow:
Steps include image preprocessing (noise removal, channel selection, vessel removal), OD size estimation, candidate region identification, accurate localization using line operator, coarse boundary detection via thresholding and ellipse fitting, and accurate boundary segmentation using level set methods.
5：Data Analysis Methods:
Accuracy metrics for localization (percentage of correct detections) and segmentation (mean distance to closest point, overlapping error), with comparisons to existing methods.
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
