Scholarlink

The brain, which is housed in the skull, is an essential organ that performs a multitude of tasks. Billion neurons work together to coordinate chemical and electrical impulses, which shape our experiences and very existence. Deep learning techniques and medical imaging have greatly increased the accuracy of early brain disease identification, including tumor and cancer detection. In medical image analysis, machine learning techniques, particularly neural-network based algorithms, have demonstrated remarkable efficacy for a range of tasks, such as the identification and categorization of cancer and brain tumors. In this paper, the CT (Computed Tomography) scans have been used for the classification of brain disease images into cancer, tumor and aneurysm. Three transfer learning (TL) models: Xception, InceptionV3 and DenseNet121 have been fine-tuned and batch normalization, dense layer and dropout layers have been added to these models. A customized Convolutional Neural Network (CNN) with 12 layers has been proposed in this paper. The fine-tuned TL models and the customized CNN have been compared and the customized CNN model executed the best and obtained an accuracy of 0.99.

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

CT-Based Brain Disease Classification Using Fine-Tuned Transfer Learning Models and Custom Convolutional Neural Network

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Preeti Gupta,
  • Swapandeep Kaur,
  • Preeti Sharma

摘要

The brain, which is housed in the skull, is an essential organ that performs a multitude of tasks. Billion neurons work together to coordinate chemical and electrical impulses, which shape our experiences and very existence. Deep learning techniques and medical imaging have greatly increased the accuracy of early brain disease identification, including tumor and cancer detection. In medical image analysis, machine learning techniques, particularly neural-network based algorithms, have demonstrated remarkable efficacy for a range of tasks, such as the identification and categorization of cancer and brain tumors. In this paper, the CT (Computed Tomography) scans have been used for the classification of brain disease images into cancer, tumor and aneurysm. Three transfer learning (TL) models: Xception, InceptionV3 and DenseNet121 have been fine-tuned and batch normalization, dense layer and dropout layers have been added to these models. A customized Convolutional Neural Network (CNN) with 12 layers has been proposed in this paper. The fine-tuned TL models and the customized CNN have been compared and the customized CNN model executed the best and obtained an accuracy of 0.99.