Related Experiment Video
Updated: Aug 24, 2025
07:23
Retinal Vascular Reactivity as Assessed by Optical Coherence Tomography Angiography
Published on: March 26, 2020
7.6K
Local-Sensitive Connectivity Filter (LS-CF): A Post-Processing Unsupervised Improvement of the Frangi, Hessian and
Erick O Rodrigues1, Lucas O Rodrigues2, João H P Machado1
1Department of Academic Informatics (DAINF), Universidade Tecnologica Federal do Parana (UTFPR), Pato Branco 85503-390, PR, Brazil.
Journal of Imaging
|October 26, 2022
Summary
This study introduces a new filter for retinal vessel analysis, improving automatic segmentation of eye blood vessels. The local-sensitive connectivity filter (LS-CF) enhances accuracy and outperforms existing methods on multiple datasets.
Area of Science:
- Ophthalmology
- Medical Imaging
- Computer Vision
Background:
- Retinal vessel analysis is crucial for assessing eye health risks.
- Accurate segmentation of retinal vessels is essential for diagnostic procedures.
- Existing methods often struggle with vessel discontinuities.
Purpose of the Study:
- To propose an unsupervised multimodal approach for improved retinal vessel segmentation.
- To introduce a novel filter, the local-sensitive connectivity filter (LS-CF), to enhance Frangi filter performance.
- To evaluate the LS-CF against existing methods and baseline approaches.
Main Methods:
- Developed the local-sensitive connectivity filter (LS-CF) incorporating pixel-level vessel continuity and a local tolerance heuristic.
- Compared LS-CF against naive connectivity filters, thresholded Frangi filter response, and morphological closing combinations.
- Evaluated performance on multimodal datasets including OSIRIX, IOSTAR, DRIVE, STARE, and CHASE-DB.
Main Results:
- The LS-CF achieved competitive results across various multimodal datasets.
- Outperformed all state-of-the-art approaches on the OSIRIX angiographic dataset for accuracy.
- Demonstrated superior performance on IOSTAR, DRIVE, STARE, and CHASE-DB datasets compared to multiple existing methods, including unsupervised ones.
Conclusions:
- The proposed LS-CF offers a robust and effective solution for unsupervised retinal vessel segmentation.
- LS-CF demonstrates significant improvements over current state-of-the-art techniques in accuracy and segmentation quality.
- This method holds promise for advancing automated risk assessment in eye care.
