DCSAU-Net: A Deeper and More Compact Split-Attention U-Net for Medical Image Segmentation

Xu, Qing; Duan, Wenting; He, Ning

doi:10.48550/arxiv.2202.00972

Cited by 7 publications

(6 citation statements)

References 15 publications

(22 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The experimental results in this work reveal that DCELANM-Net has good performance metrics on both CVC-ClinicDB and Kvasir-SEG 23,24 datasets, and that it can effectively increase the accuracy of automatic segmentation of polyp-detecting lesions from colonoscopy pictures. Moreover, the performance metrics of DCELANM-Net on the ISIC2018 35 dataset are much superior than the existing state-of-the-art models, which highlights the huge potential of DCELANM-Net in the field of medical image segmentation. Yet, the generalization ability is equally necessary for a good medical picture segmentation model.…”

Section: Discussionmentioning

confidence: 95%

“…In the ISIC2018 35 dataset, an automated skin lesion diagnostic instrument scored higher for its ability to detect melanoma, thereby increasing survival rates in many cases. In the experiment to validate DCELANM-Net's ability to generalize, we further evaluated our model using ISIC2018 data.…”

Section: Generalization To Other Datasetsmentioning

confidence: 99%

“…We discover that our model is also capable of achieving more optimal evaluation metrics with larger datasets. Our method obtains mDice of 0.912, mIou of 0.8912, and mPre of 0.9231 according to ISIC2018 35 F I G U R E 6 Each row from top to bottom reflects the experimental operations on the CVC Clinical DB, Kvasir SEG, and ISIC2018 23,24,35 datasets. Columns provide the visualization results of various models.…”

Section: Generalization To Other Datasetsmentioning

confidence: 99%

See 2 more Smart Citations

DCELANM‐Net: Medical image segmentation based on dual channel efficient layer aggregation network with learner

Lu,

Xia,

Przystupa

et al. 2023

Int J Imaging Syst Tech

View full text Add to dashboard Cite

Segmenting medical images is a principal component of computer vision. The UNet model framework has taken over as the standard framework for this activity across a wide range of medical picture segmentation applications. Due to convolutional neural networks (CNNs) convolution operation limitations, the model's global modeling ability is not absolutely perfect. Moreover, a single convolution operation cannot gather feature information at various scales, which will have an impact on the quality of the global feature extraction as well as the localization of local details. The DCELANM‐Net structure, which this article offers, is a model that ingeniously combines a Dual Channel Efficient Layer Aggregation Network (DCELAN) and a Micro Masked Autoencoder (Micro‐MAE). On the one hand, for the DCELAN, the features are more effectively fitted by deepening the network structure, which in turn can successfully learn and fuse the features, helping to locate the local feature information more accurately; and the utilization of each layer of channels is more effectively improved by widening the network structure and residual connections. We adopt Micro‐MAE as the learner of the model. In addition to being straightforward in its methodology, it also includes a self‐supervised learning method, which has the benefit of being incredibly scalable for the model. This scalable method enables the generalization of high‐volume models, and the models can show good scaling behavior. It is also shown that Micro‐MAE is a powerful and adaptable learner that we can incorporate it into our network design to improve the model's accuracy and stability for tasks involving medical picture segmentation. Superior metrics and good generalization are demonstrated by DCELANM‐Net on the datasets Kvasir‐SEG and CVC‐ClinicDB. In the experiments, we set DCELANM‐S and DCELANM‐L to represent different sizes of the model, and since DCELANM‐L has the best performance, DCELANM‐L is determined as the base model for all experiments, called DCELANM.

show abstract

Section: Discussionmentioning

confidence: 95%

Section: Generalization To Other Datasetsmentioning

confidence: 99%

Section: Generalization To Other Datasetsmentioning

confidence: 99%

See 1 more Smart Citation

DCELANM‐Net: Medical image segmentation based on dual channel efficient layer aggregation network with learner

Lu,

Xia,

Przystupa

et al. 2023

Int J Imaging Syst Tech

View full text Add to dashboard Cite

show abstract

“…Neural network design needs to consider several key indicators of network depth, width, parameters, and computation ( Bau et al, 2020 ; Lorenz, 2021d ; Xu, Duan & He, 2022 ). Because the residual structure can solve the problems of network degradation well, increasing the network depth once became the main way to improve network performance for the lack of enough computing power ( Hu et al, 2020 ; Iandola et al, 2016 ; Tan, Pang & Le, 2020 ; Wu et al, 2021 ).…”

Section: Methodsmentioning

confidence: 99%

A novel residual block: replace Conv1× 1 with Conv3×3 and stack more convolutions

Sheng

Zhang

et al. 2023

PeerJ Computer Science

View full text Add to dashboard Cite

The residual structure has an important influence on the design of the neural network model. The neural network model based on residual structure has excellent performance in computer vision tasks. However, the performance of classical residual networks is restricted by the size of receptive fields, channel information, spatial information and other factors. In this article, a novel residual structure is proposed. We modify the identity mapping and down-sampling block to get greater effective receptive field, and its excellent performance in channel information fusion and spatial feature extraction is verified by ablation studies. In order to further verify its feature extraction capability, a non-deep convolutional neural network (CNN) was designed and tested on Cifar10 and Cifar100 benchmark platforms using a naive training method. Our network model achieves better performance than other mainstream networks under the same training parameters, the accuracy we achieved is 3.08 percentage point higher than ResNet50 and 1.38 percentage points higher than ResNeXt50. Compared with SeResNet152, it is 0.29 percentage point higher in the case of 50 epochs less training.

show abstract

“…The suggested technique delivers cutting-edge results in accuracy, speed, and efficiency. Xu et al [16] proposed a novel convolutional neural network (CNN)-based structure called DCSAU-Net for various medical image segmentation applications. This framework uses multi-resolution mixed features and a broad receptive field in its CSA and DC layers.…”

Section: Literature Reviewmentioning

confidence: 99%

An Optimized Multi-Organ Cancer Cells Segmentation for Histopathological Images Based on CBAM-Residual U-Net

Shah

Kang

2023

IEEE Access

View full text Add to dashboard Cite

In digital pathology, the accurate segmentation of cell nuclei in histopathology images is essential for medical image analysis. Histopathologists visually evaluate the patterns of cellular architecture and tissue patterns in histopathology image analysis for cancer detection to determine the malignant tissue portions and assess the severity of malignancy. However, manually analyzing scans using a highresolution microscope requires significant effort and time. A computer-assisted diagnosis system utilizing deep learning (DL) algorithms rapidly, reliably, and automatically segments cell nuclei. However, the existing research studies have limited accuracy, high computational costs, and a lack of robustness and generalizability on diverse datasets. To address these issues, this paper proposes a novel and improved DL architecture based on the U-Net, namely, the CBAM-Residual U-Net for improving accuracy, robustness, and generalized segmentation algorithm that can be applied to various staining techniques and tissue structures. The proposed architecture utilizes a ResConv and convolution block attention modules (CBAM). These modules help the proposed architecture learn the image's shallow and deep features. The CBAM module uses an attention mechanism concentrating on essential features such as cell nuclei's shape, texture, and intensity to accurately segment the raw input patterns. The proposed CBAM-Residual U-Net involves fewer trainable parameters, reducing the computational and time costs compared to state-of-the-art techniques. Extensive experiments and comprehensive evaluations are conducted to demonstrate the performance of the proposed scheme on publicly available datasets: i) Data Science Bowl (DSB) 2018, ii) The GlaS, iii) Triple-Negative Breast Cancer (TNBC). The experimental results show that our proposed model considerably outperforms the state-of-the-art techniques and detects cellular boundaries well, providing fine-grained segmentation results.

show abstract

DCSAU-Net: A Deeper and More Compact Split-Attention U-Net for Medical Image Segmentation

Cited by 7 publications

References 15 publications

DCELANM‐Net: Medical image segmentation based on dual channel efficient layer aggregation network with learner

DCELANM‐Net: Medical image segmentation based on dual channel efficient layer aggregation network with learner

A novel residual block: replace Conv1× 1 with Conv3×3 and stack more convolutions

An Optimized Multi-Organ Cancer Cells Segmentation for Histopathological Images Based on CBAM-Residual U-Net

Contact Info

Product

Resources

About