Scholarlink

Brain tumor diagnosis from a Magnetic Resonant Image (MRI) serves as a critical task requiring extreme care for effective treatment of the patients. Thus, efficient and robust auto diagnosis systems are needed to identify an MRI as brain tumor affected or not. In this work, a deep semi-transfer learning approach is developed to classify an MRI as normal or any type of brain tumor, i.e., glioma, meningioma, or pituitary. Transfer Learning although effective, may not always learn features pertaining to MRIs only. Thus, a lightweight Convolutional Neural Network (CNN) is trained exclusively on MRIs, and its output is concatenated with that of a lightweight deep transfer learning model called MobileNetV2 for better learning. The concatenated output is utilized as the final feature map for classification of the input MRI. The proposed method yields a good accuracy of 97% when evaluated on the publicly available brain tumor MRI dataset.

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

Brain Tumor Classification of MRI Data Using Deep Semi-transfer Learning Framework

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Averi Ray,
  • Debam Saha,
  • Neelotpal Chakraborty

摘要

Brain tumor diagnosis from a Magnetic Resonant Image (MRI) serves as a critical task requiring extreme care for effective treatment of the patients. Thus, efficient and robust auto diagnosis systems are needed to identify an MRI as brain tumor affected or not. In this work, a deep semi-transfer learning approach is developed to classify an MRI as normal or any type of brain tumor, i.e., glioma, meningioma, or pituitary. Transfer Learning although effective, may not always learn features pertaining to MRIs only. Thus, a lightweight Convolutional Neural Network (CNN) is trained exclusively on MRIs, and its output is concatenated with that of a lightweight deep transfer learning model called MobileNetV2 for better learning. The concatenated output is utilized as the final feature map for classification of the input MRI. The proposed method yields a good accuracy of 97% when evaluated on the publicly available brain tumor MRI dataset.