Chaining a U-Net With a Residual U-Net for Retinal Blood Vessels Segmentation

Abstract

Retina images are the only non-invasive way of accessing the cardiovascular system, offering us a means of observing patterns such as microaneurysms, hemorrhages and the vasculature structure which can be used to diagnose a variety of diseases. The main goal of this paper is to automate retinal blood vessel segmentation with a good tradeoff between blood vessel classification and training time in the presence of high unbalanced classes. In this work, a novel methodology is proposed using two convolutional neural networks (CNN's), chained to each other. The second CNN has been designed with residual network blocks, which joined to the information flow from the first, give us metrics like recall and F1-Score, which are, in most cases, superior to state of the art in vessel segmentation task. We tested this work on two public datasets for blood vessel segmentation in retinal images showing that this work outperforms many of other contributions by other authors.

Retinal vessel segmentation to diagnose diabetic retinopathy using fundus images: A survey

“…Normalization and zero-phase component analysis [68,69] Contrast limited adaptive histogram equalization (CLAHE) [70][71][72][73][74][75][76][77][78][79][80][81] Gamma correction [77,82,83] Principal component analysis (PCA) [84,85] Matched and gaussian filters [86,87] Automatic color equalization [88] strategy was introduced to overcome the overfitting challenge. Since unlabeled data is used in co-training, label fusion is necessary to merge the labels obtained from cotraining to feature space to improve the results.…”

Lightweight semantic segmentation network for tumor cell nuclei and skin lesion

“…Luo et al. ( 6 ) introduce the weighted attention mechanism into the U-Net network and incorporate a Dense Connection Network ( 7 ), proposing the AD-UNet network to improve the utilization of model feature information while reducing network complexity and learning parameter complexity. Liu et al.…”

“…It also introduces a weighted attention mechanism that focuses only on the target region of interest and discards irrelevant noisy backgrounds. Luo et al (6) introduce the weighted attention mechanism into the U-Net network and incorporate a Dense Connection Network (7), proposing the AD-UNet network to improve the utilization of model feature information while reducing network complexity and learning parameter complexity. Liu et al (8) build upon the U-Net network with ResNet50 convolutional blocks and use a feature pyramid network to obtain segmentation outputs at different scales from the decoder.…”

Highly accurate blood vessel segmentation using texture‐based modified K‐means clustering with deep learning model

“…Due to the increased loss linked with the discriminator, this could segment thin arteries, although correct vasculature could be produced after performing postprocessing depending on anatomical information. For segmentation, a deep supervised FCN with bottom‐to‐top and top‐to‐bottom short interconnections was utilized 28,29 . The vascular probability map was developed by feeding blood vessel characteristics acquired to the MLP from multiscale filters including gabor and Gaussian filters.…”

“…For segmentation, a deep supervised FCN with bottom-to-top and top-to-bottom short interconnections was utilized. 28,29 The vascular probability map was developed by feeding blood vessel characteristics acquired to the MLP from multiscale filters including gabor and Gaussian filters. Many approaches and algorithms for autonomously analyzing retinal images and segmenting blood vessels 30 have been developed.…”