These tools will no longer be maintained as of December 31, 2024. Archived website can be found here. PubMed4Hh GitHub repository can be found here. Contact NLM Customer Service if you have questions.


Pubmed for Handhelds

PUBMED FOR HANDHELDS

Search MEDLINE/PubMed

  • Title: Segmenting retinal vessels with revised top-bottom-hat transformation and flattening of minimum circumscribed ellipse.
    Author: Wang W, Wang W, Hu Z.
    Journal: Med Biol Eng Comput; 2019 Jul; 57(7):1481-1496. PubMed ID: 30903529.
    Abstract:
    Retinal vessel automatic segmentation plays a great important role for analyzing fundus pathologies like diabetes, retinopathy, and hypertension. In this paper, a novel unsupervised method to automatically extract the vessels from fundus images is introduced. The method proposed a new vessel enhancement approach that we called revised top-bottom-hat transformation for removing the bright lesions for further enhancing vessels in a fundus image, and provides a novel feature that we call flattening of minimum circumscribed ellipse for recognizing a vessel. This method was tested on two publicly available databases DRIVE and STARE, and achieved an average accuracy of 0.9446 and 0.9503, respectively. For pathological cases, the approach reached an accuracy of 0.9435 and 0.9439, respectively. The time complexity approaches (O(n)), which is significantly lower than the state-of-the-art method. Graphical Abstract (GA)-Overview of the steps of the proposed algorithm Step 1: Input. Input a fundus color image. Step 2: Preprocess. The aim of process is to obtain gray image and to filter noise. Step 3: Enhancement and amendment. For improving the segmentation accuracy, a new enhancement and amendment is applied for enhancing the vessels particularly thin vessels and removing the various disturbances. Step 4: Blood vessel segmentation. Step 4.1: Binarization. To identify the blood vessel, the threshold-based method is applied to gain binary images. Step 4.2: Object decomposition. Before blood vessel recognition, we must decompose the binary image into some independent objects. Step 4.3: Calculate the flattening. Calculate the flattening of each of objects. Step 4.4: Blood vessel recognition. Blood vessels are identified by its flattening. Step 5: Output. Output a blood vessel image Graphical Abstract (GA)-Overview of the proposed approach. (a) Input Image. (b) Preprocessing. (c) Top-bottom-hat transformation. (d) Enhancement. (e) Blood vessel segmentation with different thresholds. (f) Blood vessels.
    [Abstract] [Full Text] [Related] [New Search]