Scholarlink

<p>Breast cancer is the second most common cause of death in the world for females; early accurate and reliable diagnostic methods are important. A number of deep learning models were proposed for diagnosing breast cancer, but such models have tremendous false-positive ratios and low precision, which could result in suboptimal treatment and misdiagnosis. To overcome these issues, this paper proposes a new Residual Group Channel Space Holographic Convolutional Neural Network with Pine Cone Optimization (RGCSHCNNet-PCO) for effectively identifying breast cancer from mammography and MRI images from the CBIS-DDSM and DCE-MRI datasets. To improve image quality, the method starts with pre-processing through gradient domain weighted guided image filtering. subsequently, it extracts robust features using the second-order synchroextracting transform and fast kurtrogram (SOST-FK). The classification is performed using RGCSHCNNet, which combines the strengths of Residual Group Channel and Spatial Attention (RGCSA) and Holographic Convolutional Neural Network (HCNN) for superior feature learning and representation. A feature fusion mechanism integrates multimodal data to enrich discriminative information, while Pine Cone Optimization fine-tunes model parameters for improved classification performance. With 99.97% accuracy, 99.96% precision, 99.92% F1 score, 99.93% sensitivity, and 99.94% specificity, the proposed method beats existing models and can be a highly reliable tool for computerized breast cancer diagnosis.</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

Enhanced Breast Cancer Classification Using Residual Group Channel Space Holographic CNN with Pine Cone Optimization for Mammography and MRI Imaging

  • 期刊论文

  • 发表日期：

    2025年7月11日

查看全文
  • V. Sudha,
  • Afrah Fathima Karimbanakkal Edakkattu,
  • Krishna Prakash Arunachalam,
  • Prasanna Kumar Lakineni

摘要

Breast cancer is the second most common cause of death in the world for females; early accurate and reliable diagnostic methods are important. A number of deep learning models were proposed for diagnosing breast cancer, but such models have tremendous false-positive ratios and low precision, which could result in suboptimal treatment and misdiagnosis. To overcome these issues, this paper proposes a new Residual Group Channel Space Holographic Convolutional Neural Network with Pine Cone Optimization (RGCSHCNNet-PCO) for effectively identifying breast cancer from mammography and MRI images from the CBIS-DDSM and DCE-MRI datasets. To improve image quality, the method starts with pre-processing through gradient domain weighted guided image filtering. subsequently, it extracts robust features using the second-order synchroextracting transform and fast kurtrogram (SOST-FK). The classification is performed using RGCSHCNNet, which combines the strengths of Residual Group Channel and Spatial Attention (RGCSA) and Holographic Convolutional Neural Network (HCNN) for superior feature learning and representation. A feature fusion mechanism integrates multimodal data to enrich discriminative information, while Pine Cone Optimization fine-tunes model parameters for improved classification performance. With 99.97% accuracy, 99.96% precision, 99.92% F1 score, 99.93% sensitivity, and 99.94% specificity, the proposed method beats existing models and can be a highly reliable tool for computerized breast cancer diagnosis.