A Review on Retinal Blood Vessel Enhancement and Segmentation Techniques for Color Fundus Photography

Critical Reviews in Biomedical Engineering, Journal Year: 2023, Volume and Issue: 52(1), P. 41 - 69

Published: Sept. 20, 2023

The retinal image is a trusted modality in biomedical image-based diagnosis of many ophthalmologic and cardiovascular diseases. Periodic examination the retina can help spotting these abnormalities early stage. However, to deal with today's large population, computerized analysis preferred over manual inspection. precise extraction vessel first decisive step for clinical applications. Every year, more articles are added literature that describe new algorithms problem at hand. majority review article restricted fairly small number approaches, assessment indices, databases. In this context, comprehensive different methods inevitable. It includes development first-hand classification methods. A bibliometric also presented. benefits drawbacks most commonly used techniques summarized. primary challenges, as well scope possible changes, discussed. order make fair comparison, numerous indices considered. findings survey could provide path researchers further work domain.

Language: Английский

---

Transformer-based cross-modal multi-contrast network for ophthalmic diseases diagnosis

Journal of Applied Biomedicine, Journal Year: 2023, Volume and Issue: 43(3), P. 507 - 527

Published: June 27, 2023

Language: Английский

---

What topics and emotions expressed by glaucoma patients? A sentiment analysis perspective

Social Network Analysis and Mining, Journal Year: 2024, Volume and Issue: 14(1)

Published: Aug. 8, 2024

Language: Английский

---

Segmentation of retinal vessels based on MRANet

Heliyon, Journal Year: 2022, Volume and Issue: 9(1), P. e12361 - e12361

Published: Dec. 15, 2022

The segmentation of retinal vessel takes a crucial part in computer-aided diagnosis diseases and eye disorders. However, the insufficient capillary vessels weak anti-noise interference ability make such task more difficult. To solve this problem, we proposed multi-scale residual attention network (MRANet) which is based on U-Net network. Firstly, to collect useful information about blood effectively, multi-level feature fusion block (MLF block). Then, different weights each fused are learned by using blocks, can retain while reducing redundant features. Thirdly, connection (MSR block) constructed, better extract image Finally, use DropBlock layer reduce parameters alleviate overfitting. Experiments show that DRIVE, accuracy rate AUC performance value our 0.9698 0.9899 respectively, CHASE_DB1 dataset, they 0.9755 0.9893 respectively. Our has effect compared with other methods, ensure continuity completeness segmentation.

Language: Английский

---

DPD-Net: Dual-path Proposal Discriminative Network for abnormal cell detection in cervical cytology images

Biomedical Signal Processing and Control, Journal Year: 2023, Volume and Issue: 89, P. 105887 - 105887

Published: Dec. 29, 2023

Language: Английский

---

Screening of idiopathic epiretinal membrane using fundus images combined with blood oxygen saturation and vascular morphological features

International Ophthalmology, Journal Year: 2022, Volume and Issue: 43(4), P. 1215 - 1228

Published: Oct. 7, 2022

Language: Английский

---

MIC-Net: multi-scale integrated context network for automatic retinal vessel segmentation in fundus image

Mathematical Biosciences & Engineering, Journal Year: 2023, Volume and Issue: 20(4), P. 6912 - 6931

Published: Jan. 1, 2023

<abstract><sec><title>Purpose</title><p>Accurate retinal vessel segmentation is of great value in the auxiliary screening various diseases. However, due to low contrast between ends branches fundus blood vessels and background, variable morphology optic disc cup image, task high-precision still faces difficulties. </p></sec><sec><title>Method</title><p>This paper proposes a multi-scale integrated context network, MIC-Net, which fully fuses encoder-decoder features, extracts information. First, hybrid stride sampling (HSS) block was designed encoder minimize loss helpful information caused by downsampling operation. Second, dense dilated convolution (DHDC) employed connection layer. On premise preserving feature resolution, it can perceive richer contextual Third, squeeze-and-excitation with residual connections (SERC) introduced decoder adjust channel attention adaptively. Finally, we utilized multi-layer fusion mechanism skip part, enables network consider both low-level details high-level semantic </p></sec><sec><title>Results</title><p>We evaluated proposed method on three public datasets DRIVE, STARE CHASE. In experimental results, Area under receiver operating characteristic (ROC) accuracy rate (Acc) achieved high performances 98.62%/97.02%, 98.60%/97.76% 98.73%/97.38%, respectively. </p></sec><sec><title>Conclusions</title><p>Experimental results show that obtain comparable performance compared state-of-the-art (SOTA) methods. Specifically, effectively reduce small error, thus proving promising tool for diagnosis ophthalmic diseases.</p></sec></abstract>

Language: Английский

---

Diabetic retinopathy classification based on dense connectivity and asymmetric convolutional neural network

Neural Computing and Applications, Journal Year: 2022, Volume and Issue: unknown

Published: Nov. 5, 2022

Language: Английский

---

Identification of retinal diseases based on retinal blood vessel segmentation using Dagum PDF and feature-based machine learning

The Imaging Science Journal, Journal Year: 2023, Volume and Issue: 71(5), P. 425 - 445

Published: Feb. 28, 2023

Retinal diseases is one of the major cause vision impairments worldwide and variations in fundus images help to identify these retinal diseases. Manual identification segmentation a tedious task highly prone errors. Thus, Dagum probability distribution function (PDF) based on matched filter (MF) for blood vessel proposed this work. Initially, different features are extracted from provided ensemble classifier perform disease classification. The overall performance work evaluated using popular datasets such as DRIVE, STARE, HRF, IRDiD, Kaggle retinal, ORIGA RFI. It found that specificity 0.9802 0.9838 achieved DRIVE STARE respectively. In addition, input image predicted with HRF-glaucoma '10_g.jpg', having higher prediction accuracy 95.45%.

Language: Английский

---

Ophthalmic OCT Segmentation Method Based on RCNN-Attention

IEEE Access, Journal Year: 2023, Volume and Issue: 11, P. 129601 - 129612

Published: Jan. 1, 2023

As the growth of science and technology, ophthalmic optical coherence tomography (OCT) image segmentation plays a key role in diagnosis. To improve accuracy segmentation, experiment proposed an OCT method that combines recurrent residual network (ResNet) attention mechanism (AM). In process, graph search (GS) algorithm is first used to perform fine operations on image, then convolution introduced correct phenomenon drill rapid degradation, finally theAM integrated utilization global information achieve accurate image. The results show research methodis tested different data sets, shows largest fitness value. There are four models, methods, medical technologies based full-size connection Unet 3+(UNet 3plus),recurrent convolutional neural (ResNet-RCNN) deep learning graphicssearch(DL-GS). areas under ROC curve models patient retinal layer boundary searching for automatic technology 0.956, 0.911, 0.897 0.856 respectively. When precision rate 0.900, recall rates corresponding method, 3+, ResNet-RCNN DL-GS 0.801, 0.663, 0.574 0.438 At same time, system can reach stable state within 0.501s running time method. applying practical visual ophthalmicOCTimages,the outcomes closer artificial expert annotation. Based above, method's stability better, providing certain reference optimization field.

Language: Английский

---