A lighter hybrid feature fusion framework for polyp segmentation DOI Creative Commons
Xueqiu He,

Luo Yonggang,

Min Liu

et al.

Scientific Reports, Journal Year: 2024, Volume and Issue: 14(1)

Published: Oct. 5, 2024

Colonoscopy is widely recognized as the most effective method for detection of colon polyps, which crucial early screening colorectal cancer. Polyp identification and segmentation in colonoscopy images require specialized medical knowledge are often labor-intensive expensive. Deep learning provides an intelligent efficient approach polyp segmentation. However, variability size heterogeneity boundaries interiors pose challenges accurate Currently, Transformer-based methods have become a mainstream trend these tend to overlook local details due inherent characteristics Transformer, leading inferior results. Moreover, computational burden brought by self-attention mechanisms hinders practical application models. To address issues, we propose novel CNN-Transformer hybrid model (CTHP). CTHP combines strengths CNN, excels at modeling information, global semantics, enhance accuracy. We transform computation over entire feature map into width height directions, significantly improving efficiency. Additionally, design new information propagation module introduce additional positional bias coefficients during attention process, reduces dispersal introduced deep mixed fusion Transformer. Extensive experimental results demonstrate that our proposed achieves state-of-the-art performance on multiple benchmark datasets Furthermore, cross-domain generalization experiments show exhibits excellent performance.

Language: Английский

Dynamic Frequency-Decoupled Refinement Network for Polyp Segmentation DOI Creative Commons
Yao Tong,

Jyh‐Wen Chai,

Ziqi Chen

et al.

Bioengineering, Journal Year: 2025, Volume and Issue: 12(3), P. 277 - 277

Published: March 11, 2025

Polyp segmentation is crucial for early colorectal cancer detection, but accurately delineating polyps challenging due to their variations in size, shape, and texture low contrast with surrounding tissues. Existing methods often rely solely on spatial-domain processing, which struggles separate high-frequency features (edges, textures) from low-frequency ones (global structures), leading suboptimal performance. We propose the Dynamic Frequency-Decoupled Refinement Network (DFDRNet), a novel framework that integrates frequency-domain processing. DFDRNet introduces Frequency Adaptive Decoupling (FAD) module, dynamically separates high- components, (FAR) refines these components before fusing them spatial enhance accuracy. Embedded within U-shaped encoder–decoder framework, achieves state-of-the-art performance across three benchmark datasets, demonstrating superior robustness efficiency. Our extensive evaluations ablation studies confirm effectiveness of balancing accuracy computational

Language: Английский

Citations

0
MSGAT: Multi-scale gated axial reverse attention transformer network for medical image segmentation DOI
Y.Q. Liu,

Haijiao Yun,

Yang Xia

et al.

Biomedical Signal Processing and Control, Journal Year: 2024, Volume and Issue: 95, P. 106341 - 106341

Published: April 24, 2024

Language: Английский

Citations

3
DEMF-Net: A dual encoder multi-scale feature fusion network for polyp segmentation DOI Creative Commons

Xiaorui Cao,

He Yu, Yan Kang

et al.

Biomedical Signal Processing and Control, Journal Year: 2024, Volume and Issue: 96, P. 106487 - 106487

Published: May 31, 2024

Colorectal cancer is a common malignant tumour of the gastrointestinal tract. Studies have shown that colonoscopy can be an effective screening method for detecting colon polyps and removing them to prevent development colorectal cancer. In this study, we propose new approach called Dual Encoder Multi-Scale Feature Fusion Network (DEMF-Net). This uses dual-scale Swin Transformer CNN as encoder extract semantic features at different scales. order enhance marginal characteristics irregular improve polyp detection rate, Dual-Branch Attention Module (DAF) captures shapes target through attention mechanism assigns higher weights feature channels with high contributions. Additionally, use Advanced (AFFM) establish long-range dependencies strengthen region ensure high-level are not lost. We also Characterization Supplementary Blocks (CSB) images unclear boundaries capture structure details model accuracy. conducted experiments on five widely adopted datasets showed our achieved superior results in terms both segmentation accuracy edge details.

Language: Английский

Citations

3
An Edge-Enhanced Network for Polyp Segmentation DOI Creative Commons
Yao Tong, Ziqi Chen, Zuojian Zhou

et al.

Bioengineering, Journal Year: 2024, Volume and Issue: 11(10), P. 959 - 959

Published: Sept. 25, 2024

Colorectal cancer remains a leading cause of cancer-related deaths worldwide, with early detection and removal polyps being critical in preventing disease progression. Automated polyp segmentation, particularly colonoscopy images, is challenging task due to the variability appearance low contrast between surrounding tissues. In this work, we propose an edge-enhanced network (EENet) designed address these challenges by integrating two novel modules: covariance attention (CEEA) cross-scale edge enhancement (CSEE) modules. The CEEA module leverages covariance-based enhance boundary detection, while CSEE bridges multi-scale features preserve fine-grained details. To further improve accuracy introduce hybrid loss function that combines cross-entropy edge-aware loss. Extensive experiments show EENet achieves Dice score 0.9208 IoU 0.8664 on Kvasir-SEG dataset, surpassing state-of-the-art models such as Polyp-PVT PraNet. Furthermore, it records 0.9316 0.8817 CVC-ClinicDB demonstrating its strong potential for clinical application segmentation. Ablation studies validate contribution

Language: Английский

Citations

3
Polyp segmentation in medical imaging: challenges, approaches and future directions DOI Creative Commons
Abdul Qayoom, Juanying Xie, Haider Ali

et al.

Artificial Intelligence Review, Journal Year: 2025, Volume and Issue: 58(6)

Published: March 17, 2025

Language: Английский

Citations

0
DUSFormer: Dual-Swin Transformer V2 Aggregate Network for Polyp Segmentation DOI Creative Commons
Zhangrun Xia, Jingliang Chen, Chengzhun Lu

et al.

IEEE Access, Journal Year: 2024, Volume and Issue: 12, P. 8822 - 8832

Published: Jan. 1, 2024

The convolutional neural network method has certain limitations in medical image segmentation. As a result of the limited availability polyp datasets, model framework is vulnerable to instability and overfitting during training. Beyond that, ambiguous target boundaries can make segmentation more difficult. We propose Dual-Swin Transformer V2 Aggregate Network called DUSFormer order address these issues, which be used accurately capture spatial semantic features with different complexities. Specifically, consists two encoders decoders for progressive feature extraction deep extraction. decoder uses Stepwise Feature Fusion (SFF) module locally emphasize fuse maps at various levels. This architecture enables faster efficient dissemination information all levels, enabling integration global dependencies local details. In addition, an Adaptive Correction Module (ACM) introduced construct aggregation relationship edge between layers encoder decoder, correct predictive irregular blurred boundaries, increase precision many advantages terms quantitative generalization ability on three datasets.

Language: Английский

Citations

1
Detection of Rehabilitation Training Effect of Upper Limb Movement Disorder Based on MPL-CNN DOI Creative Commons
Lijuan Shi, Runmin Wang, Jian Zhao

et al.

Sensors, Journal Year: 2024, Volume and Issue: 24(4), P. 1105 - 1105

Published: Feb. 8, 2024

Stroke represents a medical emergency and can lead to the development of movement disorders such as abnormal muscle tone, limited range motion, or abnormalities in coordination balance. In order help stroke patients recover soon possible, rehabilitation training methods employ various modes ordinary movements joint reactions induce active limbs gradually restore normal functions. Rehabilitation effect evaluation physicians understand needs different patients, determine effective treatment strategies, improve efficiency. achieve real-time accuracy action detection, this article uses Mediapipe’s detection algorithm proposes model based on MPL-CNN. Mediapipe be used identify key point features patient’s upper simultaneously hand. detect for limb disorders, LSTM CNN are combined form new LSTM-CNN model, which is extracted by Medipipe. The MPL-CNN effectively during patients. ensure scientific validity unified standards movements, employs postures Fugl-Meyer Upper Limb Training Functional Assessment Form (FMA) establishes an FMA data set experimental verification. Experimental results show that each stage MPL-CNN-based method’s recognition actions reached 95%. At same time, average rate reaches 97.54%. This shows highly robust across categories proves feasible solution. method provide high-precision effects after stroke, helping clinicians evaluating progress adjusting plan results. will personalization precision promote patient recovery.

Language: Английский

Citations

1
RSAFormer: A method of polyp segmentation with region self-attention transformer DOI
Xuehui Yin, Jun Zeng, Tianxiao Hou

et al.

Computers in Biology and Medicine, Journal Year: 2024, Volume and Issue: 172, P. 108268 - 108268

Published: March 11, 2024

Language: Английский

Citations

1
PDCA-Net: Parallel dual-channel attention network for polyp segmentation DOI
Gang Chen, Minmin Zhang,

Junmin Zhu

et al.

Biomedical Signal Processing and Control, Journal Year: 2024, Volume and Issue: 101, P. 107190 - 107190

Published: Nov. 27, 2024

Language: Английский

Citations

1
Exploring the Potential of Ensembles of Deep Learning Networks for Image Segmentation DOI Creative Commons
Loris Nanni, Alessandra Lumini, Carlo Fantozzi

et al.

Information, Journal Year: 2023, Volume and Issue: 14(12), P. 657 - 657

Published: Dec. 12, 2023

To identify objects in images, a complex set of skills is needed that includes understanding the context and being able to determine borders objects. In computer vision, this task known as semantic segmentation it involves categorizing each pixel an image. It crucial many real-world situations: for autonomous vehicles, enables identification surrounding area; medical diagnosis, enhances ability detect dangerous pathologies early, thereby reducing risk serious consequences. study, we compare performance various ensembles convolutional transformer neural networks. Ensembles can be created, e.g., by varying loss function, data augmentation method, or learning rate strategy. Our proposed ensemble, which uses simple averaging rule, demonstrates exceptional across multiple datasets. Notably, compared prior state-of-the-art methods, our ensemble consistently shows improvements well-studied polyp problem. This problem precise delineation polyps within approach showcases noteworthy advancements domain, obtaining average Dice 0.887, outperforms current SOTA with 0.885.

Language: Английский

Citations

3