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

Improving Generation and Evaluation of Long Image Sequences for Embryo Development Prediction DOI Open Access
Pedro Celard, Adrián Seara Vieira, José Manuel Sorribes-Fdez

et al.

Electronics, Journal Year: 2024, Volume and Issue: 13(3), P. 476 - 476

Published: Jan. 23, 2024

Generating synthetic time series data, such as videos, presents a formidable challenge complexity increases when it is necessary to maintain specific distribution of shown stages. One case embryonic development, where prediction and categorization are crucial for anticipating future outcomes. To address this challenge, we propose Siamese architecture based on diffusion models generate predictive long-duration development videos an evaluation method select the most realistic video in non-supervised manner. We validated model using standard metrics, Fréchet inception distance (FID), (FVD), structural similarity (SSIM), peak signal-to-noise ratio (PSNR), mean squared error (MSE). The proposed generates up 197 frames with size 128×128, considering real input images. Regarding quality all results showed improvements over default (FID = 129.18, FVD 802.46, SSIM 0.39, PSNR 28.63, MSE 97.46). On coherence stages, global stage 9.00 was achieved versus 13.31 59.3 methods. technique produces more accurate successfully removes cases that display sudden movements or changes.

Language: Английский

Citations

0
MSFE-CapsNet: a Multiscale Feature-enhanced Capsule Network for Skin Lesion Image Classification DOI
Yanjun Liu,

Haijiao Yun,

Yang Xia

et al.

Published: Jan. 19, 2024

Automatic classification of skin lesion images is an important method for learning characteristics from dermoscopic and determining the category to which they belong, crucial diagnosis treatment cancer. However, large variation in size seriously affects images. Therefore, this paper, we propose a Multiscale feature-enhanced capsule network (MSFE-CapsNet), employs 31 × convolutional kernel obtain larger perceptual domain, design multiscale feature enhancement block augment local information further learn semantic image, finally introduces Efficient Multi-Scale Attention efficiently spatial location lesions on high-level map. The experimental results show that MSFE-CapsNet achieves 95.05% accuracy HAM10000 dataset, better than existing methods classification, has only 1.58M parameters.

Language: Английский

Citations

0
A novel Parallel Cooperative Mean-Teacher framework (PCMT) combined with prediction uncertainty guide and class contrastive learning for semi-supervised polyp segmentation DOI
Yang Xia,

Haijiao Yun,

Peiyu Liu

et al.

Expert Systems with Applications, Journal Year: 2024, Volume and Issue: 255, P. 124816 - 124816

Published: July 22, 2024

Language: Английский

Citations

0
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

Citations

0