Shape-based filter for micro-aneurysm detection

https://doi.org/10.1016/j.compeleceng.2020.106620Get rights and content

Abstract

Automatic detection of micro-aneurysm in color retinal image is important for early screening and diagnosis of diabetic retinopathy. In this paper, a new method is proposed for micro-aneurysm detection based on circular bilateral Gabor filtering. Firstly, a circular bilateral Gabor filter is developed to extract micro-aneurysm candidates. Secondly, false positives are reduced by eliminating small vessels through a process involving local gradient analysis. The proposed method is tested on the retinal images from the Retinopathy Online Challenge database and Tianjin Medical University Metabolic Diseases Hospital. Evaluation results at both image and lesion level demonstrate the efficacy of the proposed method in detecting micro-aneurysm accurately.

Introduction

The incidence of diabetes is global and has serious health consequence for human population [1]. Diabetic retinopathy (DR) is one of the eye complications of diabetes that leads to decreased vision in sufferers and eventually to blindness. Micro-aneurysm (MA) is the first symptom of DR, which is a swell on the micro-vascular [2] caused by the thinning of vessel wall and blood pressure. Arguably, it is important to focus on techniques that detect micro-aneurysm as a precursor to the onset of hemorrhage and as an early diagnostic marker for DR. Furthermore, this early detection has significance for ophthalmological clinical diagnosis and precaution. The procedure of MA detection through color retinal imaging is usually performed by medical practitioners and is necessarily slow as well as demanding high human resource. Reasons adduced for this are the small sizes of MA and its random distribution in color retinal image. Consequently, development of automatic computer-aided screening systems for MA detection has attracted much attention from researchers.

The rest of the paper is organized as follows. In Section 2, short review of key existing methods in the literature is presented. The proposed method are described in Section 3. Experimental results and performance analysis are discussed in Section 4. Concluding remarks are offered in Section 5.

Section snippets

Related works

In general, MA has a quasi-circular or spherical structure that forms the basis of some structure-based methods. In a structure-based method, Lázár and Hajdu [3] used a local rotating cross-section profile to detect MA, in terms of the Gaussian-like distribution of MA. However, the results reported many false positives. They modified and improved this method by constructing an MA score map from different orientations [4]. A new criterion based on non-maximum suppression was proposed to detect

Proposed method

The proposed method can be divided into three modules, viz. preprocessing, candidate extraction and post processing, respectively. The details are shown in Fig. 1 depicting flowchart of the method.

Dataset description

Two databases are chosen to test and verify the performance of the proposed method. Retinopathy Online Challenge (ROC) [22] is firstly established to detect MA. The gold standard is annotated by four ophthalmologists. Then, the images from Tianjin Medical University Metabolic Diseases Hospital (Hospital) are collected, including 280 retinal images with 2180 × 2000 resolution at 50 field of view. Three ophthalmologists also labeled the MAs independently.

Evaluation of method and discussions

Evaluations are conducted at two

Conclusions

Micro-aneurysm is a small and dark target in retinal image. Therefore, the preprocessing algorithm of three-stage Gamma correction is used to enhance the whole lightness of the green channel image. Moreover, the local contrast of micro-aneurysm is efficient improved for post-processing. Based on the quasi-circular and Gaussian-like characteristics of micro-aneurysm, the circular bilateral Gabor filter can simultaneously improve small dark dots and suppress the surrounding complex background.

Declaration of Competing Interest

The authors declared that they have no conflicts of interest to this work We declare that we do not have any commercial or associative interest that represents a conflict of interest in connection with the work submitted

Acknowledgments

This work was supported by National Natural Science Foundation of China under Grant No. 61902078.

Xinpeng Zhang received the Ph.D degree from Tianjin Polytechnic University, China. He is currently doing his Post Doctor in Guangdong University of Technology, Guangzhou, China. His areas of interest are image processing and deep learning.

References (25)

  • C.I. Sánchez et al.

    Mixture model-based clustering and logistic regression for automatic detection of microaneurysms in retinal images

    Proceedings of the SPIE

    (2009)
  • American Diabetes Association

    Diagnosis and classification of diabetes mellitus

    Diabetes Care

    (2005)
  • C.I.O. Martins et al.

    Evaluation of retinal vessel segmentation methods for microaneurysms detection

    Proceedings of IEEE international conference on image processing

    (2009)
  • I. Lázár et al.

    Microaneurysm detection on retinal images using a rotating cross-section based model

    Proceedings of IEEE international symposium on biomedical imaging

    (2011)
  • I. Lázár et al.

    Retinal microaneurysm detection through local rotating cross-section profile analysis

    IEEE Trans Med Imaging

    (2013)
  • G. Quellec et al.

    Optimal wavelet transform for the detection of microaneurysms in retina photographs

    IEEE Trans Med Imaging

    (2008)
  • A. Bhalerao et al.

    Robust detection of microaneurysms for sight threatening retinopathy screening

    Proceedings of sixth indian conference on computer vision, graphics and image processing, Bhubaneswar, India

    (2008)
  • S. Ding et al.

    An accurate approach for microaneurysm detection in digital fundus images

    Proceedings of IEEE international conference on pattern recognition, Stockholm, Sweden

    (2014)
  • M. Niemeijer et al.

    Automatic detection of red lesions in digital color fundus photographs

    IEEE Trans. Med. Imaging

    (2005)
  • B. Zhang et al.

    Detection of microaneurysms using multi-scale correlation coefficients

    Pattern Recognit.

    (2010)
  • A.D. Fleming et al.

    Automated microaneurysm detection using local contrast normalization and local vessel detection

    IEEE Trans. Med. Imaging

    (2006)
  • L. Giancardo et al.

    Microaneurysm detection with Radon transform-based classification on retina images

    Proceedings of IEEE international conference on engineering in medicine and biology society

    (2011)
  • Cited by (4)

    Xinpeng Zhang received the Ph.D degree from Tianjin Polytechnic University, China. He is currently doing his Post Doctor in Guangdong University of Technology, Guangzhou, China. His areas of interest are image processing and deep learning.

    Zhitao Xiao received the Ph.D degree from Tianjin University, China. He is currently working as a Professor in the School of Electronics and Information Engineering, Tianjin Polytechnic University, Tianjin, China. His areas of interest include image processing, pattern recognition and computer vision.

    Fang Zhang received the Ph.D degree from Tianjin University, China. She is currently working as a Professor in the School of Electronics and Information Engineering, Tianjin Polytechnic University, Tianjin, China. Her areas of interest are image processing, pattern recognition.

    Philip O. Ogunbona received the Ph.D degree from Imperial College London, Britain. He is currently working as a Professor in the School of Electrical, Computer and Telecommunications Engineering, University of Wollongong, Australia. His areas of interest include image processing, pattern recognition and machine learning.

    Jiangtao Xi received the Ph.D degree from University of Wollongong, Australia. He is currently working as a Professor in the School of Computing and Information Technology, University of Wollongong, Australia. His areas of interest are digital signal processing and electrical engineering.

    Jun Tong received the Ph.D degree from City University of Hong Kong, HongKong. He is currently working as a Professor in the School of Computing and Information Technology, University of Wollongong, Australia. His areas of interest are signal processing.

    This paper is for CAEE special section SI-cih. Reviews processed and recommended for publication to the Editor-in-Chief by Guest Editor Dr. Hamed Vahdat-Nejad.

    View full text