A deep attention network via high-resolution representation for liver and liver tumor segmentation

Yang, Li; Zou, Beiji; Liu, Qing

doi:10.1016/j.bbe.2021.08.010

Cited by 7 publications

(4 citation statements)

References 8 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…A Dice score of 80% was achieved in this study, demonstrating that the CNN methods performed better than the classical machine learning methods AdaBoost, Random Forests, and support vector machine. Li et al [ 58 ] proposed a deep attention neural network including a high-resolution branch that can maintain input image resolution and thus preserve spatial details, as well as multiscale feature aggregation for cascading liver and tumor segmentation from CT images. This model achieved a Dice score of 76.3% for lesions and 96.0% for the liver when LiTS datasets were used for the evaluation.…”

Section: Discussionmentioning

confidence: 99%

Deep Learning Algorithm for Differentiating Patients with a Healthy Liver from Patients with Liver Lesions Based on MR Images

Skwirczyński

Tabor

Lasek

et al. 2023

Cancers

View full text Add to dashboard Cite

The problems in diagnosing the state of a vital organ such as the liver are complex and remain unresolved. These problems are underscored by frequently published studies on this issue. At the same time, demand for imaging diagnostics, preferably using a method that can detect the disease at the earliest possible stage, is constantly increasing. In this paper, we present liver diseases in the context of diagnosis, diagnostic problems, and possible elimination. We discuss the dataset and methods and present the stages of the pipeline we developed, leading to multiclass segmentation of the liver in multiparametric MR image into lesions and normal tissue. Finally, based on the processing results, each case is classified as either a healthy liver or a liver with lesions. For the training set, the AUC ROC is 0.925 (standard error 0.013 and a p-value less than 0.001), and for the test set, the AUC ROC is 0.852 (standard error 0.039 and a p-value less than 0.001). Further refinements to the proposed pipeline are also discussed. The proposed approach could be used in the detection of focal lesions in the liver and the description of liver tumors. Practical application of the developed multi-class segmentation method represents a key step toward standardizing the medical evaluation of focal lesions in the liver.

show abstract

Section: Discussionmentioning

confidence: 99%

Deep Learning Algorithm for Differentiating Patients with a Healthy Liver from Patients with Liver Lesions Based on MR Images

Skwirczyński

Tabor

Lasek

et al. 2023

Cancers

View full text Add to dashboard Cite

show abstract

“…In [23], the author proposes a model based on a self-attention module and a deep attention neural network. A high-resolution branch is also used to preserve spatial features.…”

Section: Literature Review and Related Workmentioning

confidence: 99%

The Efficacy of U-Net in Segmenting Liver Tumors from Abdominal CT Images

2022

IJIES

View full text Add to dashboard Cite

One of the primary causes of cancer-related mortality is liver cancer. Computed tomography (CT) is a commonly utilised imaging technique for the assessment and staging of hepatic tumors. Manually scanning volumetric CT scans is time intensive and ambiguous. Despite the fact that numerous deep learning models for semantic segmentation have been developed, the U-Net model remains extremely successful. In this work, we propose a method for segmenting liver tumors from abdominal CT images that is entirely based on the U-Net model and demonstrate the model's simplicity and efficacy. The U-Net architecture is employed at two levels. The first level of segmentation is used to segment the liver from CT slices, while the second level is utilised to segment tumours from masked CT images. Using CT scans from the LiTS dataset, the proposed technique achieved a dice global score of 94 percent for liver segmentation and 73 percent for tumour segmentation.

show abstract

“…HRNet learns strong highresolution representations by connecting high-and-low resolution convolutions in parallel, where there are repeated multi-scale fusions across parallel convolutions [26]. HRNet has demonstrated great segmentation and detection performance in real-world pictures such as Cityscapes and PASCAL dataset [24] as well as in medical images [27][28][29]. The nnU-Net utilizes the U-Net architecture with automatic configuration of preprocessing, network architecture, training and post-processing for any new task which known to be one of the state-of-the-art approaches for various medical image segmentation tasks.…”

Section: Introductionmentioning

confidence: 99%

3D Segmentation of Necrotic Lung Lesions in CT Images Using Self-Supervised Contrastive Learning

Liu,

Halek,

Crawford

et al. 2024

IEEE Access

View full text Add to dashboard Cite

Deep convolutional neural networks (CNN) are often trained on 2D annotations created by radiologists following RECIST guidelines to segment lesions in 3D medical images. Three-dimensional segmentation is conducted by segmenting each lesion slice-by-slice on the axial direction and stacking the 2D segmentation masks into 3D. However, the performance of such models is inherently biased by the appearance of most of the lesions in the training dataset. Herein we propose an approach to generate accurate 3D segmentations of underrepresented necrotic lung lesions. Our proposed approach applies two novel augmentation techniques for contrastive learning pretraining: dependency augmentation that captures inter-slice and inter-timepoint dependencies, and distance transform-based mask-out augmentation imitating necrotic lesions. In dependency augmentation, cosine similarity within RECIST bounding box is applied to construct positive pairs from 2D image slices of the same lesion in the current 3D volume and across longitudinal scans. We further compared contrastive learning architectures, Momentum Contrast (MoCo) and Bootstrap Your Own Latent (BYOL), based upon two internal 3D testing sets, one with regular lung lesions and the other with necrotic lung lesions, and a public 3D DeepLesion lung lesion testing set. MoCo with both proposed augmentations demonstrated the best performance among all methods that were compared. Specifically, it 1) improved Dice similarity coefficient (DSC) by 8.42% over baseline model trained from scratch and 2.40% over ImageNet pretrained model on the 3D necrotic lung lesion set; 2) achieved better segmentation performance on necrotic lesions with 10% of labeled data for supervised fine-tuning compared with the baseline model trained with all labels from scratch.

show abstract

A deep attention network via high-resolution representation for liver and liver tumor segmentation

Cited by 7 publications

References 8 publications

Deep Learning Algorithm for Differentiating Patients with a Healthy Liver from Patients with Liver Lesions Based on MR Images

Deep Learning Algorithm for Differentiating Patients with a Healthy Liver from Patients with Liver Lesions Based on MR Images

The Efficacy of U-Net in Segmenting Liver Tumors from Abdominal CT Images

3D Segmentation of Necrotic Lung Lesions in CT Images Using Self-Supervised Contrastive Learning

Contact Info

Product

Resources

About