Scholarlink

Brain tumors are a serious disorder brought on by irregular and rapidly cell division. The results of a delayed or incorrect diagnosis of a tumor might be catastrophic. Because magnetic resonance imaging (MRI) has good resolution and can clearly detect small, non-invasive abnormalities in the brain, it is a commonly utilized tool by medical professionals to diagnose brain tumors. Brain tumors rank among the leading causes of death for individuals of all ages. Accurate tumor location, size, and dimension determination is critical for successful treatment. Manual tumor classification on MRI scans takes a lot of time, even with its accuracy. To solve this issue, computerized techniques can produce more complete and precise results faster. Using “Brain Tumor MRI Dataset”, obtained from kaggle, and through the use of transfer learning techniques, this study aimed to examine the potential of five pre-trained Deep Convolutional Neural Networks (DCNNs), specifically EfficientNetB0, DenseNet121, ResNet50, InceptionV3, and Xception, CNN in the classification of brain MR data. Metrics including accuracy, recall, precision, and F1 score were used to evaluate the test set’s validation, and the results showed that the pre-trained Xception model using transfer learning performed the best out of all of these models. Furthermore, these methods offer a comprehensive classification of raw photos without requiring the human extraction of features.

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

Enhanced Brain Tumor Classification Using Transfer Learning: A Comparative Analysis of Pre-trained CNN Models on MRI Images

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Akshar Patel,
  • Khushi Shah,
  • Hemant N. Yadav

摘要

Brain tumors are a serious disorder brought on by irregular and rapidly cell division. The results of a delayed or incorrect diagnosis of a tumor might be catastrophic. Because magnetic resonance imaging (MRI) has good resolution and can clearly detect small, non-invasive abnormalities in the brain, it is a commonly utilized tool by medical professionals to diagnose brain tumors. Brain tumors rank among the leading causes of death for individuals of all ages. Accurate tumor location, size, and dimension determination is critical for successful treatment. Manual tumor classification on MRI scans takes a lot of time, even with its accuracy. To solve this issue, computerized techniques can produce more complete and precise results faster. Using “Brain Tumor MRI Dataset”, obtained from kaggle, and through the use of transfer learning techniques, this study aimed to examine the potential of five pre-trained Deep Convolutional Neural Networks (DCNNs), specifically EfficientNetB0, DenseNet121, ResNet50, InceptionV3, and Xception, CNN in the classification of brain MR data. Metrics including accuracy, recall, precision, and F1 score were used to evaluate the test set’s validation, and the results showed that the pre-trained Xception model using transfer learning performed the best out of all of these models. Furthermore, these methods offer a comprehensive classification of raw photos without requiring the human extraction of features.