Retinal vessel segmentation using a finite element based binary level set method

Zhenlin Guo,Guangrong Ji,Ping Lin,Yangfan Wang

doi:10.3934/ipi.2014.8.459

Abstract

In this paper we combine a few techniques to label blood vessels in thematched filter (MF) response image by using a finite element basedbinary level set method. An operator-splitting method is applied to numericallysolve the Euler-Lagrange equation from minimizing an energy functional.Unlike the traditional MF methods, where a threshold isdifficult to be selected, our method can automatically get more preciseblood vessel segmentation using an enhanced edge information. In order to demonstrate the good performance, we compare ourmethod with a few other methods when they are applied toa publicly available standard database of coloured images (withmanual segmentations available too).

Highlights

The main task of the medical image analysis is the extraction of an appropriate feature of the image data including organs, tissues, lesions, tumours, exudates, blood vessels etc
A variety of medical imaging techniques such as ultrasound, X-ray imaging, magnetic resonance imaging (MR), computed tomography (CT), and fluorescein angiography are capable of obtaining data on blood vessels
We have combined a few techniques for the problem of the retinal vessel segmentation

Summary

Introduction

The main task of the medical image analysis is the extraction of an appropriate feature of the image data including organs, tissues, lesions, tumours, exudates, blood vessels etc. Assessment of the characteristics of blood vessels plays an important role in medical diagnoses. A variety of medical imaging techniques such as ultrasound, X-ray imaging, magnetic resonance imaging (MR), computed tomography (CT), and fluorescein angiography are capable of obtaining data on blood vessels. In the majority of images noise and lack of contrast pose significant challenges to the extraction of blood vessels. Extracting blood vessel of medical images is even more challenging due to its large number of vessels and some of them are tiny and vague. Existing vessel segmentation methods include: rule-based methods and comprises vessel tracking [2] [3], matched filter responses [2][4][5][6][7], multithreshold probing in [8], topology adaptive snakes [9], morphology-based techniques [10][11][12], neural network approaches [13], pattern recognition and multiscale approaches [14][15][16][17][18], tight-frame methods[20][21], deformable model and level set approaches [26][27][28]

Methods

Results

Conclusion