Scholarlink

<p>Accurate diagnosis of brain tumors from MRI scans is crucial for effective treatment planning and patient outcomes. Deep learning techniques have shown promise in automating this process, but the interpretability of their decisions remains a challenge. This paper introduces the Squeeze-and-Excitation Convolutional Neural Network-MobileNet (SECNN-MNet), a novel deep learning model designed for precise and interpretable classification of brain tumors in MRI images. SECNN-MNet combines the robust feature extraction capabilities of SE-CNN with the computational efficiency of MobileNet. By integrating SE blocks, the model dynamically recalibrates feature maps to capture crucial tumor characteristics, enhancing diagnostic accuracy. MobileNet’s depthwise separable convolutions optimize computational resources while preserving image detail fidelity. Our model achieves exceptional performance, demonstrating an accuracy of 99.6% in classifying brain tumors. Leveraging explainable artificial intelligence (XAI), SECNN-MNet provides insights into decision-making processes, enhancing its interpretability and utility for clinicians. This study underscores the potential of SECNN-MNet in advancing automated brain tumor diagnosis, offering both high accuracy and interpretability crucial for clinical adoption and patient care.</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

High-precision brain tumor diagnosis using SECNN-MNet framework and explainable AI

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年5月31日

查看全文
  • Nabil M. AbdelAziz,
  • Wael Said,
  • Mohamed M. AbdelHafeez,
  • Asmaa H. Ali

摘要

Accurate diagnosis of brain tumors from MRI scans is crucial for effective treatment planning and patient outcomes. Deep learning techniques have shown promise in automating this process, but the interpretability of their decisions remains a challenge. This paper introduces the Squeeze-and-Excitation Convolutional Neural Network-MobileNet (SECNN-MNet), a novel deep learning model designed for precise and interpretable classification of brain tumors in MRI images. SECNN-MNet combines the robust feature extraction capabilities of SE-CNN with the computational efficiency of MobileNet. By integrating SE blocks, the model dynamically recalibrates feature maps to capture crucial tumor characteristics, enhancing diagnostic accuracy. MobileNet’s depthwise separable convolutions optimize computational resources while preserving image detail fidelity. Our model achieves exceptional performance, demonstrating an accuracy of 99.6% in classifying brain tumors. Leveraging explainable artificial intelligence (XAI), SECNN-MNet provides insights into decision-making processes, enhancing its interpretability and utility for clinicians. This study underscores the potential of SECNN-MNet in advancing automated brain tumor diagnosis, offering both high accuracy and interpretability crucial for clinical adoption and patient care.