Polyp-PVT: Polyp Segmentation with Pyramid Vision Transformers

arXiv (Cornell University), Год журнала: 2021, Номер unknown

Опубликована: Янв. 1, 2021

Most polyp segmentation methods use CNNs as their backbone, leading to two key issues when exchanging information between the encoder and decoder: 1) taking into account differences in contribution different-level features 2) designing an effective mechanism for fusing these features. Unlike existing CNN-based methods, we adopt a transformer encoder, which learns more powerful robust representations. In addition, considering image acquisition influence elusive properties of polyps, introduce three standard modules, including cascaded fusion module (CFM), camouflage identification (CIM), similarity aggregation (SAM). Among these, CFM is used collect semantic location polyps from high-level features; CIM applied capture disguised low-level features, SAM extends pixel area with position entire area, thereby effectively cross-level The proposed model, named Polyp-PVT, suppresses noises significantly improves expressive capabilities. Extensive experiments on five widely adopted datasets show that model various challenging situations (e.g., appearance changes, small objects, rotation) than representative methods. available at https://github.com/DengPingFan/Polyp-PVT.

Язык: Английский

---

MALUNet: A Multi-Attention and Light-weight UNet for Skin Lesion Segmentation

2021 IEEE International Conference on Bioinformatics and Biomedicine (BIBM), Год журнала: 2022, Номер unknown, С. 1150 - 1156

Опубликована: Дек. 6, 2022

Recently, some pioneering works have preferred applying more complex modules to improve segmentation performances. However, it is not friendly for actual clinical environments due limited computing resources. To address this challenge, we propose a light-weight model achieve competitive performances skin lesion at the lowest cost of parameters and computational complexity so far. Briefly, four modules: (1) DGA consists dilated convolution gated attention mechanisms extract global local feature information; (2) IEA, which based on external characterize overall datasets enhance connection between samples; (3) CAB composed 1D fully connected layers perform fusion multi-stage features generate maps channel axis; (4) SAB, operates by shared 2D spatial axis. We combine with our U-shape architecture obtain medical image dubbed as MALUNet. Compared UNet, improves mIoU DSC metrics 2.39% 1.49%, respectively, 44x 166x reduction in number complexity. In addition, conduct comparison experiments two (ISIC2017 ISIC2018). Experimental results show that achieves state-of-the-art balancing parameters, Code available https://github.com/JCruan519/MALUNet.

Язык: Английский

---

EGE-UNet: An Efficient Group Enhanced UNet for Skin Lesion Segmentation

Lecture notes in computer science, Год журнала: 2023, Номер unknown, С. 481 - 490

Опубликована: Янв. 1, 2023

Язык: Английский

---

Video Polyp Segmentation: A Deep Learning Perspective

Deleted Journal, Год журнала: 2022, Номер 19(6), С. 531 - 549

Опубликована: Ноя. 3, 2022

Abstract We present the first comprehensive video polyp segmentation (VPS) study in deep learning era. Over years, developments VPS are not moving forward with ease due to lack of a large-scale dataset fine-grained annotations. To address this issue, we introduce high-quality frame-by-frame annotated dataset, named SUN-SEG, which contains 158 690 colonoscopy frames from well-known SUN-database. provide additional annotation covering diverse types, i.e., attribute, object mask, boundary, scribble, and polygon. Second, design simple but efficient baseline, PNS+, consists global encoder, local normalized self-attention (NS) blocks. The encoders receive an anchor frame multiple successive extract long-term short-term spatial-temporal representations, then progressively refined by two NS Extensive experiments show that PNS+ achieves best performance real-time inference speed (170 fps), making it promising solution for task. Third, extensively evaluate 13 representative polyp/object models on our SUN-SEG attribute-based comparisons. Finally, discuss several open issues suggest possible research directions community. Our project publicly available at https://github.com/GewelsJI/VPS .

Язык: Английский

---

Dual encoder–decoder-based deep polyp segmentation network for colonoscopy images

Scientific Reports, Год журнала: 2023, Номер 13(1)

Опубликована: Янв. 21, 2023

Detection of colorectal polyps through colonoscopy is an essential practice in prevention cancers. However, the method itself labor intensive and subject to human error. With advent deep learning-based methodologies, specifically convolutional neural networks, opportunity improve upon prognosis potential patients suffering with cancer has appeared automated detection segmentation polyps. Polyp a number problems such as model overfitting generalization, poor definition boundary pixels, well model's ability capture practical range textures, sizes, colors. In effort address these challenges, we propose dual encoder-decoder solution named Segmentation Network (PSNet). Both encoder decoder were developed by comprehensive combination variety learning modules, including PS encoder, transformer decoder, enhanced dilated partial merge module. PSNet outperforms state-of-the-art results extensive comparative study against 5 existing polyp datasets respect both mDice mIoU at 0.863 0.797, respectively. our new modified dataset obtain 0.941 0.897

Язык: Английский

---

DuAT: Dual-Aggregation Transformer Network for Medical Image Segmentation

Lecture notes in computer science, Год журнала: 2023, Номер unknown, С. 343 - 356

Опубликована: Дек. 24, 2023

Язык: Английский

---

XBound-Former: Toward Cross-Scale Boundary Modeling in Transformers

IEEE Transactions on Medical Imaging, Год журнала: 2023, Номер 42(6), С. 1735 - 1745

Опубликована: Янв. 13, 2023

Skin lesion segmentation from dermoscopy images is of great significance in the quantitative analysis skin cancers, which yet challenging even for dermatologists due to inherent issues, i.e., considerable size, shape and color variation, ambiguous boundaries. Recent vision transformers have shown promising performance handling variation through global context modeling. Still, they not thoroughly solved problem boundaries as ignore complementary usage boundary knowledge contexts. In this paper, we propose a novel cross-scale boundary-aware transformer, XBound-Former, simultaneously address problems segmentation. XBound-Former purely attention-based network catches via three specially designed learners. First, an implicit learner (im-Bound) constrain attention on points with noticeable enhancing local modeling while maintaining context. Second, explicit (ex-Bound) extract at multiple scales convert it into embeddings explicitly. Third, based learned multi-scale embeddings, (X-Bound) by using embedding one scale guide other scales. We evaluate model two datasets polyp dataset, where our consistently outperforms convolution- transformer-based models, especially boundary-wise metrics. All resources could be found https://github.com/jcwang123/xboundformer .

Язык: Английский

---

Attention-Guided Pyramid Context Network for Polyp Segmentation in Colonoscopy Images

IEEE Transactions on Instrumentation and Measurement, Год журнала: 2023, Номер 72, С. 1 - 13

Опубликована: Янв. 1, 2023

Recently, deep convolutional neural networks (CNNs) have provided us an effective tool for automated polyp segmentation in colonoscopy images. However, most CNN-based methods do not fully consider the feature interaction among different layers and often cannot provide satisfactory performance. In this article, a novel attention-guided pyramid context network (APCNet) is proposed accurate robust Specifically, considering that represent aspects, APCNet first extracts multilayer features structure, then uses aggregation strategy to refine of each layer using complementary information layers. To obtain abundant features, extraction module (CEM) explores via local retainment global compaction. Through top-down supervision, our implements coarse-to-fine finally localizes region precisely. Extensive experiments on two in-domain four out-of-domain show comparable 19 state-of-the-art methods. Moreover, it holds more appropriate tradeoff between effectiveness computational complexity than these competing

Язык: Английский

---

BCU-Net: Bridging ConvNeXt and U-Net for medical image segmentation

Computers in Biology and Medicine, Год журнала: 2023, Номер 159, С. 106960 - 106960

Опубликована: Апрель 20, 2023

Язык: Английский

---

VM-UNET-V2: Rethinking Vision Mamba UNet for Medical Image Segmentation

Lecture notes in computer science, Год журнала: 2024, Номер unknown, С. 335 - 346

Опубликована: Янв. 1, 2024

Язык: Английский

---