Improvement in Retinal Blood Vessel Detection Using Modified Time-Frequency Analysis Algorithms fo Ophthalmic Disease Diagnosis

In medical image processing, there are significant challenges related to image clutter, uneven intensity distribution, and image blurring when extracting blood vessels. Therefore, finding an appropriate method for optimal preprocessing, noise removal, and image enhancement, as well as identifying effective filters and segmentation techniques for blood vessels, is essential. Addressing these issues can greatly assist ophthalmologists in achieving faster and easier segmentation of retinal blood vessels, thereby improving diagnostic accuracy. Therefore, accurate extraction of the retinal blood vessel tree is used for screening, eye diseases, and the diagnosis of complications resulting from cardiovascular diseases. For this purpose, a new method is employed to extract vessel centers, utilizing an improved Tophat operator combined with the application of small-sized circular structural elements. After enhancing the retinal blood vessels, the fuzzy clustering technique, specifically the C-means algorithm, is applied to estimate the actual width of the vessels. Finally, by combining the results obtained from the two stages—vessel center extraction and fuzzy clustering—the final segmentation of the blood vessels is achieved. The performance of the proposed improved algorithm has been evaluated in terms of accuracy, sensitivity, and specificity on the DRIVE and STARE datasets, and the obtained data has been analyzed. Finally, a comparison between the proposed algorithm and previous methods has been carried out , with their strengths and weaknesses analyzed and presented.