LRNet: Link Residual Neural Network for Blood Vessel Segmentation in OCTA Images

Deleted Journal, Journal Year: 2025, Volume and Issue: unknown

Published: Jan. 14, 2025

Optical coherence tomography angiography (OCTA) is an emerging, non-invasive technique increasingly utilized for retinal vasculature imaging. Analysis of OCTA images can effectively diagnose diseases, unfortunately, complex vascular structures within possess significant challenges automated segmentation. A novel, fully convolutional dense connected residual network proposed to segment the regions images. Firstly, a dual-branch structure Recurrent Residual Convolutional Neural Network (RRCNN) block constructed utilizing RecurrentBlock and operations. Subsequently, ResConvNeXt V2 Block built as backbone network. The output from then fed into side branch next Block. Within branch, Group Receptive Field (GRFB) processes results previous current layers. Ultimately, are added outputs produce final model achieves superior performance. Experiments were conducted on ROSSA OCTA-500 datasets, yielding Dice scores 91.88%, 91.72%, 89.18% respective accuracies 98.31%, 99.02%, 98.02%.

Language: Английский

---

Deep learning for retinal vessel segmentation: a systematic review of techniques and applications

Medical & Biological Engineering & Computing, Journal Year: 2025, Volume and Issue: unknown

Published: Feb. 18, 2025

Language: Английский

---

Skin lesion segmentation with a multiscale input fusion U-Net incorporating Res2-SE and pyramid dilated convolution

Scientific Reports, Journal Year: 2025, Volume and Issue: 15(1)

Published: March 7, 2025

Skin lesion segmentation is crucial for identifying and diagnosing skin diseases. Accurate aids in localizing diseases, monitoring morphological changes, extracting features further diagnosis, especially the early detection of cancer. This task challenging due to irregularity lesions dermatoscopic images, significant color variations, boundary blurring, other complexities. Artifacts like hairs, blood vessels, air bubbles complicate automatic segmentation. Inspired by U-Net its variants, this paper proposes a Multiscale Input Fusion Residual Attention Pyramid Convolution Network (MRP-UNet) dermoscopic image MRP-UNet includes three modules: Module (MIF), Res2-SE Module, Dilated (PDC). The MIF module processes different sizes morphologies fusing input information from various scales. integrates Res2Net SE mechanisms enhance multi-scale feature extraction. PDC captures at receptive fields through pyramid dilated convolution, improving accuracy. Experiments on ISIC 2016, 2017, 2018, PH2, HAM10000 datasets show that outperforms methods. Ablation studies confirm effectiveness main modules. Both quantitative qualitative analyses demonstrate MRP-UNet's superiority over state-of-the-art enhances combining multiscale fusion, residual attention, convolution. It achieves higher accuracy across multiple datasets, showing promise disease diagnosis improved patient outcomes.

Language: Английский

---

IMDF‐Net: Iterative U‐Net With Multi‐Kernel Dilated Convolution and Fusion Modules for Enhanced Retinal Vessel Segmentation

International Journal of Imaging Systems and Technology, Journal Year: 2025, Volume and Issue: 35(2)

Published: March 1, 2025

ABSTRACT In the early diagnosis of diabetic retinopathy, morphological properties blood vessels serve as an important reference for doctors to assess a patient's condition, facilitating scientific diagnostic and therapeutic interventions. However, vascular deformations, proliferation, rupture caused by retinal diseases are often difficult detect in stages. The assessment vessel morphology is subjective, time‐consuming, heavily dependent on professional experience physician. Therefore, computer‐aided systems have gradually played significant role this field. Existing neural networks, particularly U‐Net its variants, shown promising results segmentation. due information loss multiple pooling operations insufficient handling local contextual features skip connections, most segmentation methods still face challenges accurately detecting microvessels. To address these limitations assist medical staff diseases, we propose iterative network with multi‐dimensional attention multi‐scale feature fusion, named IMDF‐Net. consists backbone refinement network. network, designed cascaded multi‐kernel dilated convolution module fusion during upsampling phase. These components expand receptive field, effectively combine global features, propagate deep shallow layers. Additionally, further capture missing correct erroneous results. Experimental demonstrate that IMDF‐Net outperforms several state‐of‐the‐art DRIVE dataset, achieving best performance across all evaluation metrics. On CHASE_DB1 it achieves optimal four It demonstrates superiority both overall visual results, improvement

Language: Английский

---

A Comprehensive Review of U‐Net and Its Variants: Advances and Applications in Medical Image Segmentation

IET Image Processing, Journal Year: 2025, Volume and Issue: 19(1)

Published: Jan. 1, 2025

ABSTRACT Medical images often exhibit low and blurred contrast between lesions surrounding tissues, with considerable variation in lesion edges shapes even within the same disease, leading to significant challenges segmentation. Therefore, precise segmentation of has become an essential prerequisite for patient condition assessment formulation treatment plans. Significant achievements have been made research related U‐Net model recent years. It improves performance is extensively applied semantic medical offer technical support consistent quantitative analysis methods. First, this paper classifies image datasets on basis their imaging modalities then examines its various improvement models from perspective structural modifications. The objectives, innovative designs, limitations each approach are discussed detail. Second, we summarise four central mechanisms variant algorithms: jump‐connection mechanism, residual‐connection 3D‐UNet, transformer mechanism. Finally, examine relationships among core enhancement commonly utilized propose potential avenues strategies future advancements. This provides a systematic summary reference researchers fields, look forward designing more efficient stable network based network.

Language: Английский

---

Exploring a multi-path U-net with probability distribution attention and cascade dilated convolution for precise retinal vessel segmentation in fundus images

Scientific Reports, Journal Year: 2025, Volume and Issue: 15(1)

Published: April 18, 2025

Abstract While deep learning has become the go-to method for image denoising due to its impressive noise removal Retinal blood vessel segmentation presents several challenges, including limited labeled data, complex multi-scale structures, and susceptibility interference from lesion areas. To confront these this work offers a novel technique that integrates attention mechanisms cascaded dilated convolution module (CDCM) within multi-path U-Net architecture. First, dual-path is developed extract both coarse fine-grained structures through separate texture structural branches. A CDCM integrated gather features, enhancing model’s ability semantic features. Second, boosting algorithm incorporates probability distribution (PDA) upscaling blocks employed. This approach adjusts distribution, increasing contribution of shallow information, thereby performance in backgrounds reducing risk overfitting. Finally, output processed feature refinement module. step further refines by integrating extracting relevant Results experiments on three benchmark datasets, CHASEDB1, DRIVE, STARE, demonstrate proposed delivers improved accuracy compared existing techniques.

Language: Английский

---

Segmentation of coronary arteries from X-ray angiographic images using density based spatial clustering of applications with noise (DBSCAN)

Biomedical Signal Processing and Control, Journal Year: 2024, Volume and Issue: 101, P. 107175 - 107175

Published: Nov. 23, 2024

Language: Английский

---

Gabor-net with multi-scale hierarchical fusion of features for fundus retinal blood vessel segmentation

Journal of Applied Biomedicine, Journal Year: 2024, Volume and Issue: 44(2), P. 402 - 413

Published: April 1, 2024

Language: Английский

---

Fundus-DANet: Dilated Convolution and Fusion Attention Mechanism for Multilabel Retinal Fundus Image Classification

Applied Sciences, Journal Year: 2024, Volume and Issue: 14(18), P. 8446 - 8446

Published: Sept. 19, 2024

The difficulty of classifying retinal fundus images with one or more illnesses present missing is known as multi-lesion classification. challenges faced by current approaches include the inability to extract comparable morphological features from different lesions and resolve issue same lesion, which presents significant feature variances due grading disparities. This paper proposes a multi-disease recognition network model, Fundus-DANet, based on dilated convolution. It has two sub-modules address aforementioned issues: interclass learning module (ILM) dilated-convolution convolutional block attention (DA-CBAM). DA-CBAM uses (CBAM) convolution merge multiscale information images. ILM channel mechanism map lower dimensions, facilitating exploring latent relationships between various categories. results demonstrate that this model outperforms previous models in multilocular OIA-ODIR dataset 93% accuracy.

Language: Английский

---

Application of Artificial Intelligence for Classification, Segmentation, Early Detection, Early Diagnosis, and Grading of Diabetic Retinopathy from Fundus Retinal Images: A Comprehensive Review

IEEE Access, Journal Year: 2024, Volume and Issue: 12, P. 172499 - 172536

Published: Jan. 1, 2024

Language: Английский

---