Retinal Fundus Image Registration via Vascular Structure Graph Matching.

Kexin Deng,Xing Zhang,Xiaoqian Dai,Jie Tian,Min Xu,Jian Zheng

doi:10.1155/2010/906067

Abstract

Motivated by the observation that a retinal fundus image may contain some unique geometric structures within its vascular trees which can be utilized for feature matching, in this paper, we proposed a graph-based registration framework called GM-ICP to align pairwise retinal images. First, the retinal vessels are automatically detected and represented as vascular structure graphs. A graph matching is then performed to find global correspondences between vascular bifurcations. Finally, a revised ICP algorithm incorporating with quadratic transformation model is used at fine level to register vessel shape models. In order to eliminate the incorrect matches from global correspondence set obtained via graph matching, we proposed a structure-based sample consensus (STRUCT-SAC) algorithm. The advantages of our approach are threefold: (1) global optimum solution can be achieved with graph matching; (2) our method is invariant to linear geometric transformations; and (3) heavy local feature descriptors are not required. The effectiveness of our method is demonstrated by the experiments with 48 pairs retinal images collected from clinical patients.

Highlights

In this paper, we proposed a novel graph-based registration framework, namely graph matching (GM)-ICP, to align pairwise retinal images
Motivated by the observation that a retinal fundus image may contain some unique geometric structures within its vascular trees which can be utilized for feature matching, in this paper, we proposed a graph-based registration framework called GM-ICP to align pairwise retinal images
In order to eliminate the incorrect matches from global correspondence set obtained via graph matching, we proposed a structure-based sample consensus (STRUCT-SAC) algorithm

Summary

Introduction

We proposed a novel graph-based registration framework, namely GM-ICP, to align pairwise retinal images. In [11], a hybrid retinal image registration approach is proposed by combining both intensity-based and vessel-based methods To achieve both global and local alignments, an elastic matching scheme is used in [12] based on reconstructed vascular trees. Related work proposed recently were reported in [16, 17], where a graph transformation matching (GTM) algorithm was developed for retinal image registration and vascular characterization Their method is limited in three aspects: (1) a good initial guess of correspondences is crucial to GTM; (2) the number of graph nodes must be equal; and (3) the vessels are detected in a semi-automatic way.

Feature Extraction

The GM-ICP Algorithm

Experiments

Findings

Conclusion