Scholarlink

<p>Early detection of breast cancer is critical for improving survival rates, as it remains one of the leading causes of mortality among women worldwide. Timely and accurate diagnosis through medical imaging plays a pivotal role in effective treatment planning. Despite advancements in imaging techniques, challenges remain in the accurate segmentation of tumors from complex mammogram data, which hinders the efficiency of a breast cancer diagnosis. Existing models often struggle with the trade-off between segmentation accuracy and computational efficiency. This study proposes a deep learning-based framework that combines EfficientNet for multi-scale feature extraction and U-Net for pixel-level tumor segmentation to address these challenges. The model's performance is further enhanced by the integration of Particle Swarm Optimization (PSO), which optimizes key hyperparameters such as learning rates and batch sizes. The framework was evaluated using the CBIS-DDSM and MIAS mammogram datasets. The system achieved classification accuracies of 98.7% and 97.9%, with tumor segmentation results of 98.8%- and 98.3%-pixel accuracy. Additional performance metrics such as Intersection over Union (IoU) and Dice coefficient demonstrated superior results, with IoU scores of 91.0% and Dice coefficients exceeding 91%, outperforming standard models like ResNet50 and DenseNet. The proposed deep learning framework demonstrates high accuracy and efficiency in both breast cancer detection and tumor segmentation, offering a scalable solution for medical imaging tasks. Its robust performance suggests that the method could be applied to other medical imaging challenges beyond breast cancer detection.</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

Deep Learning Framework for Breast Cancer Detection and Segmentation Using EfficientNet and U-Net with Hyperparameter Optimization

  • 期刊论文

  • 发表日期：

    2025年6月6日

查看全文
  • Ahmed Abdel-Wahab,
  • Venkateswararao Pulipati,
  • Shubham Joshi,
  • Ramya Srikanteswara,
  • Sindhu Menon,
  • Papiya Dutta

摘要

Early detection of breast cancer is critical for improving survival rates, as it remains one of the leading causes of mortality among women worldwide. Timely and accurate diagnosis through medical imaging plays a pivotal role in effective treatment planning. Despite advancements in imaging techniques, challenges remain in the accurate segmentation of tumors from complex mammogram data, which hinders the efficiency of a breast cancer diagnosis. Existing models often struggle with the trade-off between segmentation accuracy and computational efficiency. This study proposes a deep learning-based framework that combines EfficientNet for multi-scale feature extraction and U-Net for pixel-level tumor segmentation to address these challenges. The model's performance is further enhanced by the integration of Particle Swarm Optimization (PSO), which optimizes key hyperparameters such as learning rates and batch sizes. The framework was evaluated using the CBIS-DDSM and MIAS mammogram datasets. The system achieved classification accuracies of 98.7% and 97.9%, with tumor segmentation results of 98.8%- and 98.3%-pixel accuracy. Additional performance metrics such as Intersection over Union (IoU) and Dice coefficient demonstrated superior results, with IoU scores of 91.0% and Dice coefficients exceeding 91%, outperforming standard models like ResNet50 and DenseNet. The proposed deep learning framework demonstrates high accuracy and efficiency in both breast cancer detection and tumor segmentation, offering a scalable solution for medical imaging tasks. Its robust performance suggests that the method could be applied to other medical imaging challenges beyond breast cancer detection.