Retinal Blood Vessel Segmentation Using Hybrid Features and Multi-Layer Perceptron Neural Networks

Abstract

Segmentation of retinal blood vessels is the first step for several computer aided-diagnosis systems (CAD), not only for ocular disease diagnosis such as diabetic retinopathy (DR) but also of non-ocular disease, such as hypertension, stroke and cardiovascular diseases. In this paper, a supervised learning-based method, using a multi-layer perceptron neural network and carefully selected vector of features, is proposed. In particular, for each pixel of a retinal fundus image, we construct a 24-D feature vector, encoding information on the local intensity, morphology transformation, principal moments of phase congruency, Hessian, and difference of Gaussian values. A post-processing technique depending on mathematical morphological operators is used to optimise the segmentation. Moreover, the selected feature vector succeeded in outfitting the symmetric features that provided the final blood vessel probability as a binary map image. The proposed method is tested on three known datasets: Digital Retinal Image for Extraction (DRIVE), Structure Analysis of the Retina (STARE), and CHASED_DB1 datasets. The experimental results, both visual and quantitative, testify to the robustness of the proposed method. This proposed method achieved 0.9607, 0.7542, and 0.9843 in DRIVE, 0.9632, 0.7806, and 0.9825 on STARE, 0.9577, 0.7585 and 0.9846 in CHASE_DB1, with respectable accuracy, sensitivity, and specificity performance metrics. Furthermore, they testify that the method is superior to seven similar state-of-the-art methods.

Investigation of Bilateral Similarity in Central Retinal Blood Vessels

“…The performance of our U-Net trained for segmenting CRBVs from color fundus photographs is shown in Table 3. Comparing to the approaches reported in [53]- [58], the performance of our U-Net may look worse. However, in those works, specific systems for each database were developed by using training, validation/development, and test data from the same database and by optimizing the hyperparameters specifically for each database.…”

MF2ResU-Net: A Multi-Feature Fusion Deep Learning Architecture for Retinal Blood Vessel Segmentation

“…Orlando et al [29] used fully connected conditional random field model for blood vessel segmentation, but labeling all blood vessels was a challenge. The methods proposed in recent years, such as literature [30][31][32][33][34][35][36]. Although the performance was desirable, the training procedure was complicated, although these methods reached new state-of-the-art performance in some metrics, which was not practical for real application.…”

Which Color Channel Is Better for Diagnosing Retinal Diseases Automatically in Color Fundus Photographs?

“…See Tables 2-6 for the color channel distribution in our studied previous works. G Li [123] RGB Lin [124] G Moghimirad [125] G G Badsha [130] Gr Budai [6] G Fathi [131] G Fraz [132] G Nayebifar [133] G, B Nguyen [134] G Wang [135] G G Guo [181] RGB Hu [182] RGB Jiang [183] RGB Oliveira [184] G Sangeethaa [185] G [188] RGB Lian [189] RGB Gu [190] RGB Noh [191] RGB Tan [192] RGB Wang [193] Gr Jiang [194] RGB 2020 Gao [195] RGB Feng [196] G Jin [197] RGB Tamim [198] G Sreng [199] RGB Bian [200] RGB Almubarak [201] RGB Tian [202] RGB Zhang [203] RGB Xie [204] RGB We performed all experiments using TensorFlow's Keras API 2.0.0, OpenCV 4.2.0, and Python 3.6.9. We used a standard PC with 32 GB memory, Intel 10th Gen Core i5-10400 Processor with six cores per socket, and Intel UHD Graphics 630 (CML GT2).…”