Abstract
Purpose :
Segmentation of blood vessels in retinal images is crucial for diagnosing, treating, evaluating clinical results, and early detection of retinal disorders and diseases. However, manual segmentation is arduous and complex and requires many years of specialist training. Most current research on vessel segmentation uses several steps to segment accurately. Here in, we propose a novel unsupervised automatic retinal blood vessel segmentation method based on fuzzy histogram feature enhancement with the improved Bonferroni mean-based (BM) pre-aggregation operators thereby improving the segmentation accuracy of vessels and the segmentation of small vessels.
Methods :
This work aims to propose a novel retinal vessel segmentation method consisting of two components:
Retinal blood vessel extraction by fusing color channel information by constructing the improved BM pre-aggregation operators and Enhancement of vessels appearing in the low contrast region by constructing a fuzzy histogram based on the prior feature intensity information evaluated through the improved BM pre-aggregation operator.
To check the efficacy of the proposed approach, it was evaluated on two public datasets DRIVE and STARE, respectively.
Results :
The visual representation of segmented vessels through the original fundus images is provided in Figure 1 and 2. It can be observed that the proposed approach preserves the geometrical characteristics of the vessels. The average accuracy, sensitivity, specificity, and area under the ROC curve of the proposed algorithm are 0.951, 0.763, 0.970, and 0.917 for the DRIVE database, and 0.955, 0.764, 0.971, and 0.905 for the STARE database, respectively.
Conclusions :
The proposed blood vessel segmentation method can preserve the geometrical structure of the vessels appearing in the input image by constructing the interrelationship handling BM pre-aggregation operator followed by enhancement through the fuzzy histogram method. The experimental results show the performance of the proposed approach with the fast segmentation.
This abstract was presented at the 2024 ARVO Imaging in the Eye Conference, held in Seattle, WA, May 4, 2024.