BTU-Net: Bidirectional Transformer U-Net for Buildings Segmentation DOI

Shannong Ma,

Jinsheng Fan,

Fei Ru

et al.

2021 China Automation Congress (CAC), Journal Year: 2023, Volume and Issue: unknown, P. 8880 - 8884

Published: Nov. 17, 2023

Deep convolutional neural networks have been employed in image segmentation more and recently due to their ability extract detailed properties from images. One of the most successful network frameworks for among them is encoder-decoder structure. U-Net combines an encoder a decoder segment images at pixel level tasks. uses multi-scale layers visual information; nevertheless, these are unable record long-distance correlations. In order gather both local global information about image, this work proposes bidirectional Transformer (BTU-Net) model, which draws inspiration concept. The BTU-Net structure has with five down-sampling up-sampling levels. Two-way transformer hybrid convolution modules used final three layers, whereas first two layers. With addition two-way quadratic complexity traditional self-attention mechanism decreases linearly. IoU, F1-score, accuracy, recall, precision scores our suggested model 61.9%, 67.2%, 83.9%, 63.3%, 84.3%, respectively, experiments shown that they comparable other models.

Language: Английский

MANet: a multi-attention network for automatic liver tumor segmentation in computed tomography (CT) imaging DOI Creative Commons
Kasun Gayashan Hettihewa, Thananop Kobchaisawat, Natthaporn Tanpowpong

et al.

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

Published: Nov. 16, 2023

Automatic liver tumor segmentation is a paramount important application for diagnosis and treatment planning. However, it has become highly challenging task due to the heterogeneity of shape intensity variation. capable establish diagnostic standard provide relevant radiological information all levels expertise. Recently, deep convolutional neural networks have demonstrated superiority in feature extraction learning medical image segmentation. multi-layer dense stacks make model quite inconsistent imitating visual attention awareness expertise recognition task. To bridge that capability, mechanisms developed better selection. In this paper, we propose novel network named Multi Attention Network (MANet) as fusion learn highlighting features while suppressing irrelevant The proposed followed U-Net basic architecture. Moreover, residual mechanism implemented encoder. Convolutional block module split into channel spatial modules implement encoder decoder integrated extract low-level combine with high-level ones. architecture trained evaluated on publicly available MICCAI 2017 Liver Tumor Segmentation dataset 3DIRCADb under various evaluation metrics. MANet promising results compared state-of-the-art methods comparatively small parameter overhead.

Language: Английский

Citations

13
S2DA-Net: Spatial and spectral-learning double-branch aggregation network for liver tumor segmentation in CT images DOI
Huaxiang Liu, Jie Yang, Chao Jiang

et al.

Computers in Biology and Medicine, Journal Year: 2024, Volume and Issue: 174, P. 108400 - 108400

Published: April 9, 2024

Language: Английский

Citations

4
ResTransUnet: An effective network combined with Transformer and U-Net for liver segmentation in CT scans DOI

Jiajie Ou,

Linfeng Jiang, Ting Bai

et al.

Computers in Biology and Medicine, Journal Year: 2024, Volume and Issue: 177, P. 108625 - 108625

Published: May 21, 2024

Language: Английский

Citations

4
EG-UNETR: An edge-guided liver tumor segmentation network based on cross-level interactive transformer DOI
Dongxu Cheng,

Zifang Zhou,

Jingwen Zhang

et al.

Biomedical Signal Processing and Control, Journal Year: 2024, Volume and Issue: 97, P. 106739 - 106739

Published: Aug. 10, 2024

Language: Английский

Citations

4
Efficient Liver Segmentation using Advanced 3D-DCNN Algorithm on CT Images DOI Open Access

S. Subha,

U. Kumaran

Engineering Technology & Applied Science Research, Journal Year: 2025, Volume and Issue: 15(1), P. 19324 - 19330

Published: Feb. 2, 2025

According to the latest global cancer statistics for 2022, liver ranks as ninth most common disease in women. Segmenting and distinguishing it from tumors within pose a significant challenge due complex nature of imaging. Common imaging methods such Magnetic Resonance Imaging (MRI), Computer Tomography (CT), Ultrasound (US) are employed distinguish tissue after collecting sample. Attempting partition tumor based on grayscale shades or shapes abdominal CT images is not ideal because overlapping intensity levels variability location shape soft tissues. To address this issue, study introduces an effective method image segmentation using 3D deep Convolutional Neural Network (3D-DCNN). The process involves several stages. First, undergo preprocessing enhance quality, including median filtering, adaptive converting them grayscale. feature extraction phase focuses extracting four sets features, Local Binary Pattern (LBP) Gray-Level Co-occurrence Matrix (GLCM). Additionally, Iterative Region Growing (IRG) technique developed improve Dice Similarity Coefficient (DSC) prediction by enhancing quality input obtained segmented images. This enables volumes can subsequently be used segment evaluate performance proposed 3D-DLNN approach. was implemented MATLAB, its evaluated various metrics. In experimental analysis, outperformed other methods, Jaccard with JISTS-FCM, Fuzzy C-Means (FCM), FCM Cluster Size Adjustment (FCM-CSA).

Language: Английский

Citations

0
GSA‐Net: Global Spatial Structure‐Aware Attention Network for Liver Segmentation in MR Images With Respiratory Artifacts DOI Creative Commons
Jina-tao Jiang, Dongsheng Zhou,

Muzhen He

et al.

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

Published: Jan. 1, 2025

ABSTRACT Automatic liver segmentation is of great significance for computer‐aided treatment and surgery diseases. However, respiratory motion often affects the liver, leading to image artifacts in magnetic resonance imaging (MRI) increasing difficulty. To overcome this issue, we propose a global spatial structure‐aware attention model (GSA‐Net), robust network developed difficulties caused by motion. The GSA‐Net an encoder‐decoder architecture, which extracts structure information from images identifies different objects using minimum spanning tree algorithm. network's encoder multi‐scale features with help effective lightweight channel module. decoder then transforms these bottom‐up filter modules. Combined boundary detection module, performance can be further improved. We evaluate effectiveness our method on two MRI benchmarks: one other without. Numerical evaluations benchmarks demonstrate that consistently outperforms previous state‐of‐the‐art models terms precision artifact dataset, also achieves notable results high‐quality datasets.

Language: Английский

Citations

0
BAF-Net: Bidirectional attention fusion network via CNN and transformers for the pepper leaf segmentation DOI Creative Commons
Jiangxiong Fang,

Houtao Jiang,

Shiqing Zhang

et al.

Frontiers in Plant Science, Journal Year: 2023, Volume and Issue: 14

Published: March 27, 2023

The segmentation of pepper leaves from images is great significance for the accurate control leaf diseases. To address issue, we propose a bidirectional attention fusion network combing convolution neural (CNN) and Swin Transformer, called BAF-Net, to segment image. Specially, BAF-Net first uses multi-scale feature (MSFF) branch extract long-range dependencies by constructing cascaded Transformer-based CNN-based block, which based on U-shape architecture. Then, it full-scale (FSFF) enhance boundary information attain detailed information. Finally, an adaptive module designed bridge relation MSFF FSFF features. results four datasets demonstrated that our model obtains F1 scores 96.75%, 91.10%, 97.34% 94.42%, IoU 95.68%, 86.76%, 96.12% 91.44%, respectively. Compared state-of-the-art models, proposed achieves better performance. code will be available at website: https://github.com/fangchj2002/BAF-Net.

Language: Английский

Citations

10
Semantic Segmentation of CT Liver Structures: A Systematic Review of Recent Trends and Bibliometric Analysis DOI Creative Commons
Jessica C. Delmoral, João Manuel R. S. Tavares

Journal of Medical Systems, Journal Year: 2024, Volume and Issue: 48(1)

Published: Oct. 14, 2024

Abstract The use of artificial intelligence (AI) in the segmentation liver structures medical images has become a popular research focus past half-decade. performance AI tools screening for this task may vary widely and been tested literature various datasets. However, no scientometric report provided systematic overview scientific area. This article presents bibliometric review recent advances neuronal network modeling approaches, mainly deep learning, to outline multiple directions field terms algorithmic features. Therefore, detailed most relevant publications addressing fully automatic semantic segmenting Computed Tomography (CT) algorithm objective, benchmark, model complexity is provided. suggests that hybrid 2D 3D networks are top performers liver. In case tumor vasculature segmentation, generative approaches perform best. reported benchmark indicates there still much be improved such small high-resolution abdominal CT scans.

Language: Английский

Citations

3
Study on photofluorescent uranium ore sorting based on deep learning DOI

Jun Qiu,

Yan Zhang, Chunqing Fu

et al.

Minerals Engineering, Journal Year: 2023, Volume and Issue: 206, P. 108523 - 108523

Published: Dec. 10, 2023

Language: Английский

Citations

8
ViT-SAPS: Detail-Aware Transformer for Mechanical Assembly Semantic Segmentation DOI Creative Commons
Haitao Dong, Chengjun Chen, Jinlei Wang

et al.

IEEE Access, Journal Year: 2023, Volume and Issue: 11, P. 41467 - 41479

Published: Jan. 1, 2023

Semantic segmentation of mechanical assembly images provides an effective way to monitor the process and improve product quality. Compared with other deep learning models, Transformer has advantages in modeling global context, it been widely applied various computer vision tasks including semantic segmentation. However, pays same granularity attention on all regions image, so some difficulty be images, which parts have large size differences information quantity distribution is uneven. This paper proposes a novel Transformer-based model called Vision Self-Adaptive Patch Size (ViT-SAPS). ViT-SAPS can perceive detail image finer-grained where locates, thus meeting requirements Specifically, self-adaptive patch splitting algorithm proposed split into patches sizes. The more region has, smaller into. Further, handle these unfixed-size patches, position encoding scheme non-uniform bilinear interpolation used after sequence decoding are proposed. Experimental results show that stronger ability than fixed size, achieves impressive locality-globality trade-off. study not only practical method for segmentation, but also much value application Transformers fields. code available at: https://github.com/QDLGARIM/ViT-SAPS.

Language: Английский

Citations

2