Scholarlink

<p>Abnormal growths of malignant or non-malignant tissue in the brain can lead to long-term damage. One important area of research in medical imaging focuses on brain tumor classification, which plays a key role in diagnosis and helps guide effective treatment planning. This paper proposes a novel Explainable Hybrid Le-GhostNet for Brain Tumor Classification (EHLGN-BTC) framework. The brain tumor classification framework starts with input MRI images, which are augmented to increase dataset size and improve model performance. The images are then pre-processed using Adaptive Dual-based Median Filtering (AD-MF) to reduce noise while preserving key details. Tumor regions are segmented using a Modified Backbone Region-based Convolutional Neural Network (MBR-CNN) and the segmented areas are classified using Hybrid Le-GhostNet (HL-GNet) model with a combination of Convolution with Global Hierarchical-based LeNet (CGH-LNet) and GhostNet models. The system produces the tumor type as output and explainable AI using LIME (XAI-LIME) highlights the regions influencing the prediction and ensures both accurate and interpretable results. The EHLGN-BTC framework incorporates a hybrid classifier (HL-GNet) for brain tumor classification, achieving a precision of 90% with 80% of the training data and an accuracy of 94% with 90% of the training data.</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

Explainable AI with Hybrid Deep Learning Model for Brain Tumor Classification

  • 期刊论文

  • 发表日期：

    2026年5月26日

查看全文
  • Guruvammal S,
  • Narmada B,
  • Agashini V Kumar

摘要

Abnormal growths of malignant or non-malignant tissue in the brain can lead to long-term damage. One important area of research in medical imaging focuses on brain tumor classification, which plays a key role in diagnosis and helps guide effective treatment planning. This paper proposes a novel Explainable Hybrid Le-GhostNet for Brain Tumor Classification (EHLGN-BTC) framework. The brain tumor classification framework starts with input MRI images, which are augmented to increase dataset size and improve model performance. The images are then pre-processed using Adaptive Dual-based Median Filtering (AD-MF) to reduce noise while preserving key details. Tumor regions are segmented using a Modified Backbone Region-based Convolutional Neural Network (MBR-CNN) and the segmented areas are classified using Hybrid Le-GhostNet (HL-GNet) model with a combination of Convolution with Global Hierarchical-based LeNet (CGH-LNet) and GhostNet models. The system produces the tumor type as output and explainable AI using LIME (XAI-LIME) highlights the regions influencing the prediction and ensures both accurate and interpretable results. The EHLGN-BTC framework incorporates a hybrid classifier (HL-GNet) for brain tumor classification, achieving a precision of 90% with 80% of the training data and an accuracy of 94% with 90% of the training data.