Scholarlink

<p>Breast cancer (BC) is a preliminary effect of breast tissues that leads to the premature death of women in worldwide. Accurate and automatic detection is complex in the field of medical image processing. In this paper, a novel deep learning-related approach is proposed named as Deep Residual Caps Network (DRCN) to detect BC. Here, different sources such as the Breast Histopathology Images dataset, Breast Cancer Histopathological (BreakHis) dataset and BACH-ICAR-2018 dataset are used to collect the histopathological (HP) images. The gathered HP images are preprocessed by using different kinds of preprocessing and segmentation approaches for enhancing diversity, reducing overfitting issues and minimizing class imbalance issues. The detection task is performed by combining Sparse Graph Convolution Residual Network (SGCResNet) and Shared Capsule Network (SCapsNet) modules. The SGCResNet module is used to learn the features for generating high-quality features by enhancing the generalization ability of the model. After that, the SCapsNet module is applied to mimic the hierarchical relationships of the model during BC detection. To enhance the operation speed of the model, the Capsule Filter Routing (CFR) is implemented which helps to filter capsules according to its activation values. The experiments are used to find the superior performance of the proposed model through different analyses such as comparison study, visual representation and numerical evaluation with diverse measures. The results showed that the proposed model achieved better performances of 98.82% and 0.94 from detection accuracy and Mathew’s Correlation Coefficient (MCC) compared to other methods.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Integrating sparse graph convolution and capsule networks for superior breast cancer diagnosis

  • 期刊论文

  • 发表日期：

    2025年3月13日

查看全文
  • P. Manju Bala,
  • U. Palani

摘要

Breast cancer (BC) is a preliminary effect of breast tissues that leads to the premature death of women in worldwide. Accurate and automatic detection is complex in the field of medical image processing. In this paper, a novel deep learning-related approach is proposed named as Deep Residual Caps Network (DRCN) to detect BC. Here, different sources such as the Breast Histopathology Images dataset, Breast Cancer Histopathological (BreakHis) dataset and BACH-ICAR-2018 dataset are used to collect the histopathological (HP) images. The gathered HP images are preprocessed by using different kinds of preprocessing and segmentation approaches for enhancing diversity, reducing overfitting issues and minimizing class imbalance issues. The detection task is performed by combining Sparse Graph Convolution Residual Network (SGCResNet) and Shared Capsule Network (SCapsNet) modules. The SGCResNet module is used to learn the features for generating high-quality features by enhancing the generalization ability of the model. After that, the SCapsNet module is applied to mimic the hierarchical relationships of the model during BC detection. To enhance the operation speed of the model, the Capsule Filter Routing (CFR) is implemented which helps to filter capsules according to its activation values. The experiments are used to find the superior performance of the proposed model through different analyses such as comparison study, visual representation and numerical evaluation with diverse measures. The results showed that the proposed model achieved better performances of 98.82% and 0.94 from detection accuracy and Mathew’s Correlation Coefficient (MCC) compared to other methods.