Scholarlink

In the field of precision medicine, distinguishing cancerous tissues within Whole Slide Images (WSIs) is critical for early diagnosis and individualized treatment planning. This study introduces a deep learning framework to classify lung, breast, and kidney tumors using WSIs from The Cancer Genome Atlas (TCGA). State-of-the-art convolutional architectures, including CNNs, KimiaNet, and EfficientNet, are utilized for reliable feature extraction and preprocessing. The framework incorporates advanced segmentation techniques, such as color-based and deep learning-assisted segmentation, to isolate tumor regions, reducing background noise and improving classification accuracy. Quantitative evaluation using accuracy, F1-score, precision, and recall metrics demonstrates superior performance, with CNN achieving an overall accuracy of 99.0%, surpassing KimiaNet (81.37%) and EfficientNet (97.0%). The proposed framework showcases the potential for optimizing digital pathology workflows, supporting pathologists with high-precision automated analysis, and paving the way for future advancements in AI-driven clinical diagnostics.

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-Based Classification of Lung, Kidney, and Breast Cancer Tumor Tissues Using Whole Slide Images from the TCGA Database

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • I. R. Oviya,
  • Korrayi Saiteja,
  • M. Harshitha,
  • Anagha Rajan

摘要

In the field of precision medicine, distinguishing cancerous tissues within Whole Slide Images (WSIs) is critical for early diagnosis and individualized treatment planning. This study introduces a deep learning framework to classify lung, breast, and kidney tumors using WSIs from The Cancer Genome Atlas (TCGA). State-of-the-art convolutional architectures, including CNNs, KimiaNet, and EfficientNet, are utilized for reliable feature extraction and preprocessing. The framework incorporates advanced segmentation techniques, such as color-based and deep learning-assisted segmentation, to isolate tumor regions, reducing background noise and improving classification accuracy. Quantitative evaluation using accuracy, F1-score, precision, and recall metrics demonstrates superior performance, with CNN achieving an overall accuracy of 99.0%, surpassing KimiaNet (81.37%) and EfficientNet (97.0%). The proposed framework showcases the potential for optimizing digital pathology workflows, supporting pathologists with high-precision automated analysis, and paving the way for future advancements in AI-driven clinical diagnostics.