Scholarlink

<p>Deep learning, particularly Convolutional Neural Networks (CNNs), has become an essential tool for histopathological image analysis, offering the ability to automatically extract complex features from medical images. In the context of ovarian cancer, accurate subtype classification is crucial for effective treatment planning and improving patient outcomes. However, similarities among subtypes and the limited size of annotated datasets pose significant challenges.</p><p>This study investigates the application of CNNs to ovarian cancer subtype classification using histopathological images. We employed data augmentation strategies, stain normalization, and transfer learning to address class imbalance, small dataset size, and staining variability. Six CNN architectures; ResNet-50, ResNet-50V2, ResNet-101, VGG-16, VGG-19, and InceptionResNetV3 were evaluated. Among these, ResNet-50V2 achieved the best performance with an accuracy of 84.37%, along with balanced precision, recall, and F1-scores.</p><p>An ablation study was conducted to evaluate the contribution of different augmentation techniques. Furthermore, a human–machine comparison was performed to assess the model’s clinical applicability. The results demonstrate that CNN-based models, when supported by appropriate preprocessing techniques, can assist in improving ovarian cancer diagnosis and subtype classification.</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

Ovarian cancer subtype classification using convolutional neural networks: an evaluation of deep learning techniques for histopathological image analysis

  • 期刊论文

  • 发表日期：

    2025年10月4日

查看全文
  • Noha Youssef,
  • Katia Gabriel,
  • Laila El Saadawi,
  • Farida Simaika,
  • Hana Issa

摘要

Deep learning, particularly Convolutional Neural Networks (CNNs), has become an essential tool for histopathological image analysis, offering the ability to automatically extract complex features from medical images. In the context of ovarian cancer, accurate subtype classification is crucial for effective treatment planning and improving patient outcomes. However, similarities among subtypes and the limited size of annotated datasets pose significant challenges.

This study investigates the application of CNNs to ovarian cancer subtype classification using histopathological images. We employed data augmentation strategies, stain normalization, and transfer learning to address class imbalance, small dataset size, and staining variability. Six CNN architectures; ResNet-50, ResNet-50V2, ResNet-101, VGG-16, VGG-19, and InceptionResNetV3 were evaluated. Among these, ResNet-50V2 achieved the best performance with an accuracy of 84.37%, along with balanced precision, recall, and F1-scores.

An ablation study was conducted to evaluate the contribution of different augmentation techniques. Furthermore, a human–machine comparison was performed to assess the model’s clinical applicability. The results demonstrate that CNN-based models, when supported by appropriate preprocessing techniques, can assist in improving ovarian cancer diagnosis and subtype classification.