Synthesizing Images With Annotations for Medical Image Segmentation Using Diffusion Probabilistic Model DOI Creative Commons

Zengan Huang,

Qinzhu Yang, Mu Tian

et al.

International Journal of Imaging Systems and Technology, Journal Year: 2024, Volume and Issue: 35(1)

Published: Dec. 14, 2024

ABSTRACT To alleviate the burden of manual annotation, there are numerous excellent segmentation models for images being developed. However, performance these data‐driven is frequently constrained by availability samples sizes pair medical and annotations. Therefore, to address this challenge, study introduces image augmentation diffusion model (MEDSAD). MEDSAD solves problem annotation scarcity utilizing a given simple generate paired images. improve stability, we used traditional study. exert better control over texture synthesis in generated MEDSAD, style injection (TSI) mechanism introduced. Additionally, propose feature frequency domain attention (FFDA) module mitigate adverse effects high‐frequency noise during generation. The efficacy substantiated through validation three distinct tasks encompassing magnetic resonance (MR) ultrasound (US) imaging modalities, focusing on breast tumors, brain nerve structures. findings demonstrate model's proficiency synthesizing pairs based provided annotations, thereby facilitating notable subsequent tasks. Moreover, improvement becomes greater as quantity synthetic available data increases. This underscores robust generalization capability intrinsic model, potentially offering avenues future explorations training research.

Language: Английский

SSTrans-Net: Smart Swin Transformer Network for medical image segmentation DOI

Liyao Fu,

Yunzhu Chen, Wei Ji

et al.

Biomedical Signal Processing and Control, Journal Year: 2024, Volume and Issue: 91, P. 106071 - 106071

Published: Feb. 2, 2024

Language: Английский

Citations

21
DPF-Net: A Dual-Path Progressive Fusion Network for Retinal Vessel Segmentation DOI
Jianyong Li, Ge Gao, Lei Yang

et al.

IEEE Transactions on Instrumentation and Measurement, Journal Year: 2023, Volume and Issue: 72, P. 1 - 17

Published: Jan. 1, 2023

Precise segmentation of retinal vessels from fundus images is essential for intervention in numerous diseases, and helpful preventing treating blindness. Deep convolutional neural network (DCNN) based approaches have achieved an excellent success the automatic vessels. However, a single (CNN) structure can only capture limited local features lack ability to extract global contexts. Meanwhile, strategies used feature fusion low-level detail information with high-level semantic fail handle phenomenon gap issue between encoder decoder validly. Therefore, high-precision still remains challenging task. In this paper, dual-path progressive network, named DPF-Net, proposed accurate end-to-end images. To detect rich formation, effective representation, which contains CNN path detecting recurrent extracting contextual information. It could acquire sufficient detailed at same time. addition, strategy aggregation scale, adjacent scales all scales, composed by interactive (IF) block, cross-layer (CLF) block scale (SFF) block. Combine maps different paths IF fuse obtain features. CLF guide representation through Finally, SFF recalculate weights realize scales. Extensive experiments conducted on three publicly available datasets (DRIVE, CHASEDB1 STARE). Experimental results show that DPF-Net achieve better compared other state-of-the-art methods, especially indeed promotes significantly boosts performance.

Language: Английский

Citations

23
CoVi-Net: A hybrid convolutional and vision transformer neural network for retinal vessel segmentation DOI
Minshan Jiang,

Yongfei Zhu,

Xuedian Zhang

et al.

Computers in Biology and Medicine, Journal Year: 2024, Volume and Issue: 170, P. 108047 - 108047

Published: Jan. 29, 2024

Language: Английский

Citations

12
A Cross-scale Attention-Based U-Net for Breast Ultrasound Image Segmentation DOI

Teng Wang,

Jun Liu, Jinshan Tang

et al.

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

Published: Jan. 21, 2025

Breast cancer remains a significant global health concern and is leading cause of mortality among women. The accuracy breast diagnosis can be greatly improved with the assistance automatic segmentation ultrasound images. Research has demonstrated effectiveness convolutional neural networks (CNNs) transformers in segmenting these Some studies combine CNNs, using transformer's ability to exploit long-distance dependencies address limitations inherent networks. Many face due forced integration transformer blocks into CNN architectures. This approach often leads inconsistencies feature extraction process, ultimately resulting suboptimal performance for complex task medical image segmentation. paper presents CSAU-Net, cross-scale attention-guided U-Net, which combined CNN-transformer structure that leverages local detail depiction CNNs handle dependencies. To integrate context data, we propose cross-attention block embedded within skip connections U-shaped architectural network. further enhance incorporated gated dilated convolution (GDC) module lightweight channel self-attention (LCAT) on encoder side. Extensive experiments conducted three open-source datasets demonstrate our CSAU-Net surpasses state-of-the-art techniques lesions.

Language: Английский

Citations

1
Biomedical image segmentation algorithm based on dense atrous convolution DOI Creative Commons

Hong'an Li,

Man Liu, Jiangwen Fan

et al.

Mathematical Biosciences & Engineering, Journal Year: 2024, Volume and Issue: 21(3), P. 4351 - 4369

Published: Jan. 1, 2024

<abstract><p>Biomedical images have complex tissue structures, and there are great differences between of the same part different individuals. Although deep learning methods made some progress in automatic segmentation biomedical images, accuracy is relatively low for with significant changes targets, also problems missegmentation missed segmentation. To address these challenges, we proposed a image method based on dense atrous convolution. First, added convolution module (DAC) encoding decoding paths U-Net network. This was inception structure design, which can effectively capture multi-scale features images. Second, introduced residual pooling to detect by connecting blocks sizes. Finally, network, adopted an attention mechanism suppress background interference enhancing weight target area. These modules work together improve robustness The experimental results showed that compared mainstream networks, our model exhibited stronger ability when processing multiple-shaped targets. At time, this significantly reduce phenomenon missegmentation, accuracy, make closer real situation.</p></abstract>

Language: Английский

Citations

6
Wavelet scattering transform application in classification of retinal abnormalities using OCT images DOI Creative Commons
Zahra Baharlouei, Hossein Rabbani, Gerlind Plonka

et al.

Scientific Reports, Journal Year: 2023, Volume and Issue: 13(1)

Published: Nov. 3, 2023

To assist ophthalmologists in diagnosing retinal abnormalities, Computer Aided Diagnosis has played a significant role. In this paper, particular Convolutional Neural Network based on Wavelet Scattering Transform (WST) is used to detect one four abnormalities from Optical Coherence Tomography (OCT) images. Predefined wavelet filters network decrease the computation complexity and processing time compared deep learning methods. We use two layers of WST obtain direct efficient model. generates sparse representation images which translation-invariant stable concerning local deformations. Next, Principal Component Analysis classifies extracted features. evaluate model using publicly available datasets have comprehensive comparison with literature. The accuracies classifying OCT OCTID dataset into five classes were [Formula: see text] text], respectively. achieved an accuracy detecting Diabetic Macular Edema Normal ones TOPCON device-based dataset. Heidelberg Duke contain DME, Age-related Degeneration, classes, we A our results state-of-the-art models shows that outperforms these for some assessments or achieves nearly best reported so far while having much smaller computational complexity.

Language: Английский

Citations

12
Multi-scale local-global transformer with contrastive learning for biomarkers segmentation in retinal OCT images DOI
Xiaoming Liu,

Yuanzhe Ding,

Ying Zhang

et al.

Journal of Applied Biomedicine, Journal Year: 2024, Volume and Issue: 44(1), P. 231 - 246

Published: Jan. 1, 2024

Language: Английский

Citations

4
Direction-guided network for retinal vessel segmentation in OCTA images DOI
Zhenli Li, Xinpeng Zhang, Meng Zhao

et al.

Biomedical Signal Processing and Control, Journal Year: 2025, Volume and Issue: 103, P. 107455 - 107455

Published: Jan. 8, 2025

Language: Английский

Citations

0
SAM-OCTA: Prompting segment-anything for OCTA image segmentation DOI
Xinrun Chen, Chengliang Wang,

Haojian Ning

et al.

Biomedical Signal Processing and Control, Journal Year: 2025, Volume and Issue: 106, P. 107698 - 107698

Published: Feb. 21, 2025

Language: Английский

Citations

0
SVVD-NET: A framework with relative position constraints for vertebral vertex detection DOI
Yongkang Xu,

Lianhong Duan,

Zhicheng Zhang

et al.

Biomedical Signal Processing and Control, Journal Year: 2025, Volume and Issue: 106, P. 107746 - 107746

Published: March 3, 2025

Language: Английский

Citations

0