A Densely Connected Network Based on U-Net for Medical Image Segmentation

Yang, Zhenzhen; Xu, Pengfei; Yang, Yongpeng

doi:10.1145/3446618

Cited by 16 publications

(4 citation statements)

References 26 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Meanwhile, the integration of residual blocks [8] not only solved the problem of gradient vanishing or exploding but also enhanced the training stability and performance of the model. Advanced versions such as Unet++ [20] and DenseUNet [21] further pushed the development of the model. Unet++ effectively dealt with challenges of multi-scale information and class imbalance by combining the advantages of U-Net and residual networks [22], while DenseUNet utilized dense blocks for more robust feature extraction and integrated multiple feature fusion technologies to improve the precision of feature extraction.…”

Section: Based On Cnn Architecturementioning

confidence: 99%

Pocket Convolution Mamba for Brain Tumor Segmentation

Zhang,

Zhao,

Wang

et al. 2024

Preprint

View full text Add to dashboard Cite

In the field of brain tumor segmentation, models based on CNNs and Transformer have received a lot of attention. However, CNNs have limitations in long-range modeling, and although Transformers can model at long distances, they have quadratic computational complexity. Recently, State Space Models (SSM), exemplified by the Mamba model, can achieve linear computational complexity and are adept at long-distance interactions. In this paper, we propose Pocket Convolution Mamba (P-BTS), which utilizes the PocketNet paradigm, SSM, and patch contrastive learning to achieve an efficient segmentation model. Specifically, the encoder follows the PocketNet paradigm, the SSM is used at the highest level of the model encoder to capture rich semantic information, and patch contrastive learning is achieved through the results of dual-stream data. Meanwhile, we designed a Spatial Channel Attention (SCA) module to enhance control over spatial channels, and a Feature Complement Module (FCM) to facilitate the interaction between low-level features and high-level semantic information. We conducted comprehensive experiments on the BraTS2018 and BraTS2019 datasets, and the results show that P-BTS has excellent segmentation performance.

show abstract

Section: Based On Cnn Architecturementioning

confidence: 99%

Pocket Convolution Mamba for Brain Tumor Segmentation

Zhang,

Zhao,

Wang

et al. 2024

Preprint

View full text Add to dashboard Cite

show abstract

“…On the other hand, the high-level features have stronger semantic features with lower perception details and poor spatial information. Therefore, fusing dense blocks with different levels by using MFF block was proposed [49]. Dense-Net uses the same concept of the identity connections as Res-Net, but with the difference that each layer receives the feature maps from all the preceding layers.…”

Section: Dense U-netmentioning

confidence: 99%

U-Net-Based Models towards Optimal MR Brain Image Segmentation

et al. 2023

View full text Add to dashboard Cite

Brain tumor segmentation from MRIs has always been a challenging task for radiologists, therefore, an automatic and generalized system to address this task is needed. Among all other deep learning techniques used in medical imaging, U-Net-based variants are the most used models found in the literature to segment medical images with respect to different modalities. Therefore, the goal of this paper is to examine the numerous advancements and innovations in the U-Net architecture, as well as recent trends, with the aim of highlighting the ongoing potential of U-Net being used to better the performance of brain tumor segmentation. Furthermore, we provide a quantitative comparison of different U-Net architectures to highlight the performance and the evolution of this network from an optimization perspective. In addition to that, we have experimented with four U-Net architectures (3D U-Net, Attention U-Net, R2 Attention U-Net, and modified 3D U-Net) on the BraTS 2020 dataset for brain tumor segmentation to provide a better overview of this architecture’s performance in terms of Dice score and Hausdorff distance 95%. Finally, we analyze the limitations and challenges of medical image analysis to provide a critical discussion about the importance of developing new architectures in terms of optimization.

show abstract

“…In [10], researchers proposed the dense UNet network and compared it with the multiresolution U-Net (multi-ResUNet) and conventional UNet networks on three separate datasets. The testing results reveal that the dense UNet network outperforms the multi-ResUNet and conventional UNet networks by a wide margin.…”

Section: Introductionmentioning

confidence: 99%

Deep Neural Network for Lung Image Segmentation on Chest X-ray

et al. 2022

View full text Add to dashboard Cite

COVID-19 patients require effective diagnostic methods, which are currently in short supply. In this study, we explained how to accurately identify the lung regions on the X-ray scans of such people’s lungs. Images from X-rays or CT scans are critical in the healthcare business. Image data categorization and segmentation algorithms have been developed to help doctors save time and reduce manual errors during the diagnosis. Over time, CNNs have consistently outperformed other image segmentation algorithms. Various architectures are presently based on CNNs such as ResNet, U-Net, VGG-16, etc. This paper merged the U-Net image segmentation and ResNet feature extraction networks to construct the ResUNet++ network. The paper’s novelty lies in the detailed discussion and implementation of the ResUNet++ architecture in lung image segmentation. In this research paper, we compared the ResUNet++ architecture with two other popular segmentation architectures. The ResNet residual block helps us in lowering the feature reduction issues. ResUNet++ performed well compared with the UNet and ResNet architectures by achieving high evaluation scores with the validation dice coefficient (96.36%), validation mean IoU (94.17%), and validation binary accuracy (98.07%). The novelty of this research paper lies in a detailed discussion of the UNet and ResUNet architectures and the implementation of ResUNet++ in lung images. As per our knowledge, until now, the ResUNet++ architecture has not been performed on lung image segmentation. We ran both the UNet and ResNet models for the same amount of epochs and found that the ResUNet++ architecture achieved higher accuracy with fewer epochs. In addition, the ResUNet model gave us higher accuracy (94%) than the UNet model (92%).

show abstract

A Densely Connected Network Based on U-Net for Medical Image Segmentation

Cited by 16 publications

References 26 publications

Pocket Convolution Mamba for Brain Tumor Segmentation

Pocket Convolution Mamba for Brain Tumor Segmentation

U-Net-Based Models towards Optimal MR Brain Image Segmentation

Deep Neural Network for Lung Image Segmentation on Chest X-ray

Contact Info

Product

Resources

About