Scholarlink

<p>Brain tumor classification is essential for accurate diagnosis and treatment planning, significantly enhancing patient outcomes and survival rates. The complexity of multi-class classification, which includes Glioma, Meningioma, Pituitary tumors, and healthy cases (No Tumor), presents considerable challenges. This study addresses these challenges through advanced preprocessing techniques such as resizing, normalization, object-centric image extraction, and contrast enhancement. Additionally, a segregation approach is employed, processing MRI images from three perspectives-Axial, Sagittal, and Coronal-to further improve classification accuracy. A modified VGG-16 model, enhanced with transfer learning, is used to extract meaningful features while optimizing the number of trainable parameters for improved efficiency. The proposed approach demonstrates impressive results, achieving 98.5% accuracy and strong performance metrics, including an Area Under the Curve of 99.0%, recall of 98.4%, specificity of 99.4%, precision of 98.3%, and an F1-score of 98.3%. These outcomes underscore the potential of the proposed method in aiding medical professionals to make more informed decisions, ultimately improving patient care and clinical efficacy.</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

Computer-aided diagnosis for multi-class classification of brain tumors using CNN features via transfer-learning

  • 期刊论文

  • 发表日期：

    2025年3月15日

查看全文
  • Agnesh Chandra Yadav,
  • Krish Shah,
  • Aaryan Purohit,
  • Maheshkumar H. Kolekar

摘要

Brain tumor classification is essential for accurate diagnosis and treatment planning, significantly enhancing patient outcomes and survival rates. The complexity of multi-class classification, which includes Glioma, Meningioma, Pituitary tumors, and healthy cases (No Tumor), presents considerable challenges. This study addresses these challenges through advanced preprocessing techniques such as resizing, normalization, object-centric image extraction, and contrast enhancement. Additionally, a segregation approach is employed, processing MRI images from three perspectives-Axial, Sagittal, and Coronal-to further improve classification accuracy. A modified VGG-16 model, enhanced with transfer learning, is used to extract meaningful features while optimizing the number of trainable parameters for improved efficiency. The proposed approach demonstrates impressive results, achieving 98.5% accuracy and strong performance metrics, including an Area Under the Curve of 99.0%, recall of 98.4%, specificity of 99.4%, precision of 98.3%, and an F1-score of 98.3%. These outcomes underscore the potential of the proposed method in aiding medical professionals to make more informed decisions, ultimately improving patient care and clinical efficacy.