Multi-Stage Attentive Network for Motion Deblurring via Binary Cross-Entropy Loss

Guo, Cai; Chen, Xinan; Chen, Yanhua; Yu, Chuying

doi:10.3390/e24101414

Cited by 13 publications

(5 citation statements)

References 36 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The model uses a binary cross-entropy loss function 15 to compute the loss, which is generally used to measure the difference between two probability distributions, and in particular to assess the discrepancy between the probabilities predicted by the model and the actual target. As shown in formula ( 4 ): where w and h are the dimensions of the image, corresponds to the pixels in the segmented image.…”

Section: Results and Analysismentioning

confidence: 99%

Segmentation of liver CT images based on weighted medical transformer model

Gu,

Zhang,

Cai

et al. 2024

Sci Rep

View full text Add to dashboard Cite

Deep convolutional neural networks have made significant strides in the field of medical image segmentation. Although existing convolutional structures enhance performance by leveraging local image information, they often lose the interdependence information between contexts. Therefore, the article utilizes the multi-attention mechanism of the Transformer structure to more comprehensively express relationships between contexts and introduced the Transformer network architecture into the field of medical image segmentation. Most models based on this Transformer structure typically require large datasets for training. However, in the medical field, the limited size of datasets makes training models with the Transformer structure challenging. To address this, the article propose a Weighted Medical Transformer (WMT) model that imposes low requirements on dataset quantity. The weighting mechanism in the WMT model aims to improve the issue of inaccurate relative positional coding when dealing with small medical datasets. Additionally, a coarse-grained and fine-grained segmentation mechanism is introduced, focusing on both the detailed aspects within image blocks and the boundary information connecting blocks. Experimental results on a liver dataset demonstrate that the model achieves F1 and IoU scores of 88.48% and 79.41%, respectively. Results on the MoNuSeg dataset show comparable high F1 and IoU scores of 79.58% and 66.19%, respectively. The model’s accuracy surpasses that of U-Net++ and U-Net models. Compared to other models, this approach is applicable to scenarios with limited datasets, exhibiting high execution efficiency and accuracy.

show abstract

Section: Results and Analysismentioning

confidence: 99%

Segmentation of liver CT images based on weighted medical transformer model

Gu,

Zhang,

Cai

et al. 2024

Sci Rep

View full text Add to dashboard Cite

show abstract

“…The diagnostic model was constructed based on the feedforward neural network algorithm to investigate whether the collective interaction of 7 hub genes can effectively distinguish breast cancer patients from normal individuals. The binary cross-entropy loss function was used as the model’s loss function when constructing the diagnostic model [ 25 ]. The formula for calculating this function was: …”

Section: Methodsmentioning

confidence: 99%

Stratification and prognostic evaluation of breast cancer subtypes defined by obesity-associated genes

Chen,

Xie,

Yang

et al. 2024

Discov Onc

View full text Add to dashboard Cite

Objective Breast cancer was the most common type of cancer among women worldwide, significantly impacting their quality of life and survival rates. And obesity has been widely accepted as an important risk factor for breast cancer. However, the specific mechanisms by which obesity affects breast cancer were still unclear. Therefore, studying the impact mechanisms of obesity as a risk factor for breast cancer was of utmost importance. Methods This study was based on TCGA breast cancer RNA transcriptomic data and the GeneCard obesity gene set. Through single and multiple factor Cox analysis and LASSO coefficient screening, seven hub genes were identified. The independent mechanisms of these seven hub genes were evaluated from various aspects, including survival data, genetic mutation data, single-cell sequencing data, and immune cell data. Additionally, the risk prognosis model and the neural network diagnostic model were established to further investigate these seven hub genes. In order to achieve precision treatment for breast cancer (BRCA), based on the RNA transcriptomic data of the seven genes, 1226 BRCA patients were divided into two subtypes: BRCA subtype 1 and BRCA subtype 2. By studying and comparing the immune microenvironment, investigating the mechanisms of differential gene expression, and exploring the mechanisms of subnetworks, we aim to explore the clinical differences in the presentation of BRCA subtypes and achieve precision treatment for BRCA. Finally, qRT-PCR experiments were conducted to validate the conclusions of the bioinformatics analysis. Results The 7 hub genes showed good diagnostic independence and can serve as excellent biomarkers for molecular diagnosis. However, they do not perform well as independent prognostic molecular markers for BRCA patients. When predicting the survival of BRCA patients, their AUC values at 1 year, 3 years, and 5 years are mostly below 0.5. Nevertheless, through the establishment of the risk prognosis model considering the combined effect of the seven hub genes, it was found that the survival prediction of BRCA patients can be significantly improved. The risk prognosis model, compared to the independent use of the seven hub genes as prognostic markers, achieved higher timeROC AUC values at 1 year, 3 years, and 5 years, with values of 0.651, 0.669, and 0.641 respectively. Additionally, the neural network diagnostic model constructed from the 7 genes performs well in diagnosing BRCA, with an AUC value of 0.94, accurately identifying BRCA patients. The two subtypes identified by the seven hub genes exhibited significant differences in survival period, with subtype 1 having a poor prognosis. The differential mechanisms between the two subtypes mainly originate from regulatory differences in the immune microenvironment. Finally, the results of this study’s bioinformatics analysis were validated through qRT-PCR experiments. Conclusion 7 hub genes serve as excellent independent biomarkers for molecular diagnosis, and the neural network diagnostic model can accurately distinguish BRCA patients. In addition, based on the expression levels of these seven genes in BRCA patients, two subtypes can be reliably identified: BRCA subtype 1 and BRCA subtype 2, and these two subtypes showed significant differences in BRCA patient survival prognosis, proportion of immune cells, and expression levels of immune cells. Among them, patients with subtype 1 of BRCA had a poor prognosis.

show abstract

“…To achieve improved convergence during training, a gradual decrease in learning rate based on the number of rounds was employed. To ensure model performance, two loss functions were used: the Complete Intersection over Union (CIoU) [16] loss function for bounding box [17] and the Binary Cross-Entropy (BCE) [18] loss function for target existence and classification. Clear and concise language was used throughout, with causal connections between statements.…”

Section: Implementation Detailsmentioning

confidence: 99%

Enhancing mask detection performance based on YOLOv5 model optimization and attention mechanisms

Yang

2024

ACE

View full text Add to dashboard Cite

Due to the COVID-19 pandemic, there has been a significant increase in the usage of masks, leading to more complex scenarios for mask detection techniques. This paper focuses on optimizing the performance of mask detection using the You Only Look Once (YOLO) v5 model. In this study, the yolov5 target detection model was employed for training the mask dataset. Diverse model improvement techniques were explored to enhance the model's capability to capture crucial features and differentiate masks from the background in complex scenarios. Finally, the modified model was compared with the earlier original target detection model to identify the most considerable performance gain. The CSPDarknet design with the TensorFlow framework is utilized in this study, and the Attention Mechanism module is implemented through the Keras library. The objective is to optimize the three feature layers between the backbone network and the neck by integrating multiple attention mechanisms. This will enable the model to more quickly and accurately capture important features when dealing with complex scenarios by adjusting the feature map weights. Additionally, in the feature pyramid network, shallow feature maps are fused with deeper feature maps in a certain order to determine the most efficient feature fusion method. Finally, this study identified the optimal combination of attention mechanism and feature fusion through ablation experiments. The results of the experiment demonstrate that the combination of SE block and shallow feature fusion (SE + FF2 model) can greatly enhance category confidence, leading to an improved model performance.

show abstract

Multi-Stage Attentive Network for Motion Deblurring via Binary Cross-Entropy Loss

Cited by 13 publications

References 36 publications

Segmentation of liver CT images based on weighted medical transformer model

Segmentation of liver CT images based on weighted medical transformer model

Stratification and prognostic evaluation of breast cancer subtypes defined by obesity-associated genes

Enhancing mask detection performance based on YOLOv5 model optimization and attention mechanisms

Contact Info

Product

Resources

About