Scholarlink

In recent years, deep learning (DL) techniques, particularly convolutional neural networks (CNNs), have emerged as powerful tools for cancer classification across various imaging modalities. This paper shows a unified CNN architecture for classifying multiple cancers. It uses transfer learning from EfficientNetB0 along with advanced convolutional techniques, such as grouped and depthwise separable convolutions. The model was trained and evaluated on eight different cancer types, achieving accuracy rates ranging from 89.4% to 97.5%. A key strength of the proposed approach lies in its ability to generalize across diverse cancer types while maintaining computational efficiency. The model’s performance was also checked using precision, recall, F1-score, and accuracy metrics. This showed how well it dealt with problems like unequal classes and different dataset sizes. The results show that the model works well for classifying multiple cancers and point out areas where more work needs to be done, like making the model work better on cancer types that aren’t used very often and making it more general across different imaging modalities.

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

Efficient Multi-cancer Detection: A Unified CNN Approach Leveraging Transfer Learning and Depthwise Convolutions

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Jagan Mohan Dudala

摘要

In recent years, deep learning (DL) techniques, particularly convolutional neural networks (CNNs), have emerged as powerful tools for cancer classification across various imaging modalities. This paper shows a unified CNN architecture for classifying multiple cancers. It uses transfer learning from EfficientNetB0 along with advanced convolutional techniques, such as grouped and depthwise separable convolutions. The model was trained and evaluated on eight different cancer types, achieving accuracy rates ranging from 89.4% to 97.5%. A key strength of the proposed approach lies in its ability to generalize across diverse cancer types while maintaining computational efficiency. The model’s performance was also checked using precision, recall, F1-score, and accuracy metrics. This showed how well it dealt with problems like unequal classes and different dataset sizes. The results show that the model works well for classifying multiple cancers and point out areas where more work needs to be done, like making the model work better on cancer types that aren’t used very often and making it more general across different imaging modalities.