Scholarlink

<p>Breast cancer has recently overtaken cervical cancer as the predominant cancer type in Indian urban areas. Despite considerable research and the development of automated diagnostic machines, current methods are far from perfect, necessitating more reliable medical assessments. Moreover, there's been relatively little research on Indian datasets compared to international resources, despite significant differences due to factors like denser breasts, varying textures, lesion sizes, and compositions. To address this gap, our work focuses on using machine learning to automate breast cancer detection, tailoring models to Indian breast types and utilizing metadata for more accurate assessments. Our analyses show promising results, with the best single model achieving a per-image AUC of 0.95, and averaging four models increasing AUC to 0.98 (sensitivity: 86.7%, specificity: 96.1%) on an independent test set from the INbreast database. These findings highlight the potential of deep learning techniques in improving mammography accuracy and reducing false positives and negatives in clinical settings.</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

Optimized Machine Learning Techniques for Precise Breast Cancer Detection in Mammograms

  • 期刊论文

  • 发表日期：

    2025年4月15日

查看全文
  • Puttegowda Kiran,
  • V. Veeraprathap,
  • U. Rajashekhar,
  • Mathapati Mahantesh,
  • K. V. Sudheesh,
  • Kumara Bharath,
  • K. Prabhavathi

摘要

Breast cancer has recently overtaken cervical cancer as the predominant cancer type in Indian urban areas. Despite considerable research and the development of automated diagnostic machines, current methods are far from perfect, necessitating more reliable medical assessments. Moreover, there's been relatively little research on Indian datasets compared to international resources, despite significant differences due to factors like denser breasts, varying textures, lesion sizes, and compositions. To address this gap, our work focuses on using machine learning to automate breast cancer detection, tailoring models to Indian breast types and utilizing metadata for more accurate assessments. Our analyses show promising results, with the best single model achieving a per-image AUC of 0.95, and averaging four models increasing AUC to 0.98 (sensitivity: 86.7%, specificity: 96.1%) on an independent test set from the INbreast database. These findings highlight the potential of deep learning techniques in improving mammography accuracy and reducing false positives and negatives in clinical settings.