LUVS-Net: A Lightweight U-Net Vessel Segmentor for Retinal Vasculature Detection in Fundus Images

Muhammad Talha Islam; Haroon Ahmed Khan; Khuram Naveed; Ali Nauman; Sardar Muhammad Gulfam; Sung Won Kim

doi:10.3390/electronics12081786

LUVS-Net: A Lightweight U-Net Vessel Segmentor for Retinal Vasculature Detection in Fundus Images

Muhammad Talha Islam, Haroon Ahmed Khan, Khuram Naveed, Ali Nauman, Sardar Muhammad Gulfam, Sung Won Kim^*

^*Corresponding author for this work

Department of Electrical and Computer Engineering - Signal Processing and Machine learning

Research output: Contribution to journal/Conference contribution in journal/Contribution to newspaper › Journal article › Research › peer-review

7 Citations (Scopus)

8 Downloads (Pure)

Abstract

This paper presents LUVS-Net, which is a lightweight convolutional network for retinal vessel segmentation in fundus images that is designed for resource-constrained devices that are typically unable to meet the computational requirements of large neural networks. The computational challenges arise due to low-quality retinal images, wide variance in image acquisition conditions and disparities in intensity. Consequently, the training of existing segmentation methods requires a multitude of trainable parameters for the training of networks, resulting in computational complexity. The proposed Lightweight U-Net for Vessel Segmentation Network (LUVS-Net) can achieve high segmentation performance with only a few trainable parameters. This network uses an encoder–decoder framework in which edge data are transposed from the first layers of the encoder to the last layer of the decoder, massively improving the convergence latency. Additionally, LUVS-Net’s design allows for a dual-stream information flow both inside as well as outside of the encoder–decoder pair. The network width is enhanced using group convolutions, which allow the network to learn a larger number of low- and intermediate-level features. Spatial information loss is minimized using skip connections, and class imbalances are mitigated using dice loss for pixel-wise classification. The performance of the proposed network is evaluated on the publicly available retinal blood vessel datasets DRIVE, CHASE_DB1 and STARE. LUVS-Net proves to be quite competitive, outperforming alternative state-of-the-art segmentation methods and achieving comparable accuracy using trainable parameters that are reduced by two to three orders of magnitude compared with those of comparative state-of-the-art methods.

Original language	English
Article number	1786
Journal	Electronics
Volume	12
Issue	8
Number of pages	16
ISSN	2079-9292
DOIs	https://doi.org/10.3390/electronics12081786
Publication status	Published - Apr 2023

Keywords

CHASE
convolutional neural network
deep learning
DRIVE
image segmentation
medical images
STARE
U-Net vessel segmentation

Access to Document

10.3390/electronics12081786Licence: CC BY

electronics-12-01786Final published version, 3.36 MBLicence: CC BY

Cite this

@article{f890a9f4aed04c538b3be0924de340a9,

title = "LUVS-Net: A Lightweight U-Net Vessel Segmentor for Retinal Vasculature Detection in Fundus Images",

abstract = "This paper presents LUVS-Net, which is a lightweight convolutional network for retinal vessel segmentation in fundus images that is designed for resource-constrained devices that are typically unable to meet the computational requirements of large neural networks. The computational challenges arise due to low-quality retinal images, wide variance in image acquisition conditions and disparities in intensity. Consequently, the training of existing segmentation methods requires a multitude of trainable parameters for the training of networks, resulting in computational complexity. The proposed Lightweight U-Net for Vessel Segmentation Network (LUVS-Net) can achieve high segmentation performance with only a few trainable parameters. This network uses an encoder–decoder framework in which edge data are transposed from the first layers of the encoder to the last layer of the decoder, massively improving the convergence latency. Additionally, LUVS-Net{\textquoteright}s design allows for a dual-stream information flow both inside as well as outside of the encoder–decoder pair. The network width is enhanced using group convolutions, which allow the network to learn a larger number of low- and intermediate-level features. Spatial information loss is minimized using skip connections, and class imbalances are mitigated using dice loss for pixel-wise classification. The performance of the proposed network is evaluated on the publicly available retinal blood vessel datasets DRIVE, CHASE_DB1 and STARE. LUVS-Net proves to be quite competitive, outperforming alternative state-of-the-art segmentation methods and achieving comparable accuracy using trainable parameters that are reduced by two to three orders of magnitude compared with those of comparative state-of-the-art methods.",

keywords = "CHASE, convolutional neural network, deep learning, DRIVE, image segmentation, medical images, STARE, U-Net vessel segmentation",

author = "Islam, {Muhammad Talha} and Khan, {Haroon Ahmed} and Khuram Naveed and Ali Nauman and Gulfam, {Sardar Muhammad} and Kim, {Sung Won}",

year = "2023",

month = apr,

doi = "10.3390/electronics12081786",

language = "English",

volume = "12",

journal = "Electronics",

issn = "2079-9292",

publisher = "Multidisciplinary Digital Publishing Institute (MDPI)",

number = "8",

}

TY - JOUR

T1 - LUVS-Net

T2 - A Lightweight U-Net Vessel Segmentor for Retinal Vasculature Detection in Fundus Images

AU - Islam, Muhammad Talha

AU - Khan, Haroon Ahmed

AU - Naveed, Khuram

AU - Nauman, Ali

AU - Gulfam, Sardar Muhammad

AU - Kim, Sung Won

PY - 2023/4

Y1 - 2023/4

N2 - This paper presents LUVS-Net, which is a lightweight convolutional network for retinal vessel segmentation in fundus images that is designed for resource-constrained devices that are typically unable to meet the computational requirements of large neural networks. The computational challenges arise due to low-quality retinal images, wide variance in image acquisition conditions and disparities in intensity. Consequently, the training of existing segmentation methods requires a multitude of trainable parameters for the training of networks, resulting in computational complexity. The proposed Lightweight U-Net for Vessel Segmentation Network (LUVS-Net) can achieve high segmentation performance with only a few trainable parameters. This network uses an encoder–decoder framework in which edge data are transposed from the first layers of the encoder to the last layer of the decoder, massively improving the convergence latency. Additionally, LUVS-Net’s design allows for a dual-stream information flow both inside as well as outside of the encoder–decoder pair. The network width is enhanced using group convolutions, which allow the network to learn a larger number of low- and intermediate-level features. Spatial information loss is minimized using skip connections, and class imbalances are mitigated using dice loss for pixel-wise classification. The performance of the proposed network is evaluated on the publicly available retinal blood vessel datasets DRIVE, CHASE_DB1 and STARE. LUVS-Net proves to be quite competitive, outperforming alternative state-of-the-art segmentation methods and achieving comparable accuracy using trainable parameters that are reduced by two to three orders of magnitude compared with those of comparative state-of-the-art methods.

AB - This paper presents LUVS-Net, which is a lightweight convolutional network for retinal vessel segmentation in fundus images that is designed for resource-constrained devices that are typically unable to meet the computational requirements of large neural networks. The computational challenges arise due to low-quality retinal images, wide variance in image acquisition conditions and disparities in intensity. Consequently, the training of existing segmentation methods requires a multitude of trainable parameters for the training of networks, resulting in computational complexity. The proposed Lightweight U-Net for Vessel Segmentation Network (LUVS-Net) can achieve high segmentation performance with only a few trainable parameters. This network uses an encoder–decoder framework in which edge data are transposed from the first layers of the encoder to the last layer of the decoder, massively improving the convergence latency. Additionally, LUVS-Net’s design allows for a dual-stream information flow both inside as well as outside of the encoder–decoder pair. The network width is enhanced using group convolutions, which allow the network to learn a larger number of low- and intermediate-level features. Spatial information loss is minimized using skip connections, and class imbalances are mitigated using dice loss for pixel-wise classification. The performance of the proposed network is evaluated on the publicly available retinal blood vessel datasets DRIVE, CHASE_DB1 and STARE. LUVS-Net proves to be quite competitive, outperforming alternative state-of-the-art segmentation methods and achieving comparable accuracy using trainable parameters that are reduced by two to three orders of magnitude compared with those of comparative state-of-the-art methods.

KW - CHASE

KW - convolutional neural network

KW - deep learning

KW - DRIVE

KW - image segmentation

KW - medical images

KW - STARE

KW - U-Net vessel segmentation

UR - http://www.scopus.com/inward/record.url?scp=85156256833&partnerID=8YFLogxK

U2 - 10.3390/electronics12081786

DO - 10.3390/electronics12081786

M3 - Journal article

AN - SCOPUS:85156256833

SN - 2079-9292

VL - 12

JO - Electronics

JF - Electronics

IS - 8

M1 - 1786

ER -

LUVS-Net: A Lightweight U-Net Vessel Segmentor for Retinal Vasculature Detection in Fundus Images

Abstract

Keywords

Access to Document

Other files and links

Fingerprint

Cite this