Abstract
Background: Fundus image is a projection of the inner surface of the eye, which can be used to analyze and judge the distribution of blood vessels on the retina due to its different shape, bifurcation and elongation. Vascular trees are the most stable features in medical images and can be used for biometrics. Ophthalmologists can effectively screen and determine the ophthalmic conditions of diabetic retinopathy, glaucoma and microaneurysms by the morphology of blood vessels presented in the fundus images. Traditional unsupervised learning methods include matched filtering method, morphological processing method, deformation model method, etc. However, due to the great difference in the feature complexity of different fundus image morphology, the traditional methods are relatively simple in coding, poor in the extraction degree of vascular features, poor in segmentation effect, and unable to meet the needs of practical clinical assistance. Methods: In this paper, we propose a new feature fusion model based on non-subsampled shearwave transform for retinal blood vessel segmentation. The contrast between blood vessels and background is enhanced by pre-processing. The vascular contour features and detailed features are extracted under the multi-scale framework, and then the image is postprocessed. The fundus images are decomposed into low frequency sub-band and high frequency sub-band by non-subsampled shear-wave transform. The two feature images are fused by regional definition weighting and guided filtering respectively, and the vascular detection image is obtained by calculating the maximum value of the corresponding pixels at each scale. Finally, the Otsu method is used for segmentation. Results: The experimental results on DRIVE data set show that the proposed method can accurately segment the vascular contour while retaining a large number of small vascular branches with high accuracy. Conclusion: The proposed method has a high accuracy and can perform vascular segmentation well on the premise of ensuring sensitivity.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.