Scholarlink

Brain tumor diagnosis via MRI is challenging due to the brain’s complexity and tumor variability. This research investigates transfer learning techniques by fine-tuning pre-trained models (Xception, InceptionV3, EfficientNetV2S) for brain tumor classification into glioma, meningioma, pituitary tumors, and no tumor. Through experiments on MRI datasets (Kaggle, Figshare, SARTAJ, Br35H), preprocessing steps like normalization and augmentation were applied to enhance data quality. Integrating Gradient-weighted Class Activation Mapping (Grad-CAM) ensures model interpretability. Experimental results reveal high classification accuracy, with Xception achieving 99%, showcasing the potential for AI-driven diagnostic tools in healthcare.

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

Innovative Transfer Learning Technique for Brain Tumor Diagnosis in Medical Resonance Imaging

  • 章节

  • 发表日期：

    2026年

查看全文
  • Naufal Nazaruddin,
  • Muhammad Zarlis

摘要

Brain tumor diagnosis via MRI is challenging due to the brain’s complexity and tumor variability. This research investigates transfer learning techniques by fine-tuning pre-trained models (Xception, InceptionV3, EfficientNetV2S) for brain tumor classification into glioma, meningioma, pituitary tumors, and no tumor. Through experiments on MRI datasets (Kaggle, Figshare, SARTAJ, Br35H), preprocessing steps like normalization and augmentation were applied to enhance data quality. Integrating Gradient-weighted Class Activation Mapping (Grad-CAM) ensures model interpretability. Experimental results reveal high classification accuracy, with Xception achieving 99%, showcasing the potential for AI-driven diagnostic tools in healthcare.