Scholarlink

<p>Brain tumors are a major cause of mortality worldwide, making early detection essential for timely intervention and improved patient prognosis. Moreover, the varying and irregular growth patterns of brain tumor tissues present challenges for clinicians in accurately identifying tumors. Although numerous deep learning techniques have been utilized in prior research, they have not reached sufficient accuracy in detecting tumors. Additionally, conventional methods are labor-intensive, leading to high healthcare costs. To address these challenges, this manuscript proposes an effective multi-class categorization and attributes selection method for brain tumor recognition using Fast Point Dual-Channel Attention-Based Convolutional Neural Networks (FaPDCACNNet). The input MRI brain images are sourced from the Figshare, BRATS2018, and Brain MRI datasets. To begin, an Adaptive Tri-Plateau Limit Tri-Histogram Equalization Algorithm (ATri-PLTri-HEA) is introduced for preprocessing, aiming to improve the quality of the input MRI images and eliminate noise. Then the preprocessed MRI images are segmented using the Masked-attention Mask Transformer (Mask2Former) technique, extracts features through the Fast Point Transformer (FPT), and classified with the Dual-Channel Convolutional Neural Network with Attention-Pooling (DCCNNet-AP) method, which is optimized using the Exponential Distribution Optimizer (EDO). The proposed FaPDCACNNet method is implemented in Python. This approach achieves 98.5% accuracy on three datasets, outperforming existing methods by enhancing feature selection and multi-class classification, making it a trustworthy and precise diagnostic tool for brain tumors.</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

Multi-Class Classification and Feature Selection-Based Brain Tumor Detection Using Fast Point Dual-Channel Attention-Based Convolutional Neural Networks

  • 期刊论文

  • 发表日期：

    2025年6月9日

查看全文
  • Diksha Dani,
  • Gaurav Agrawal,
  • Angappan Kumaresan,
  • Jackulin Thangarasu

摘要

Brain tumors are a major cause of mortality worldwide, making early detection essential for timely intervention and improved patient prognosis. Moreover, the varying and irregular growth patterns of brain tumor tissues present challenges for clinicians in accurately identifying tumors. Although numerous deep learning techniques have been utilized in prior research, they have not reached sufficient accuracy in detecting tumors. Additionally, conventional methods are labor-intensive, leading to high healthcare costs. To address these challenges, this manuscript proposes an effective multi-class categorization and attributes selection method for brain tumor recognition using Fast Point Dual-Channel Attention-Based Convolutional Neural Networks (FaPDCACNNet). The input MRI brain images are sourced from the Figshare, BRATS2018, and Brain MRI datasets. To begin, an Adaptive Tri-Plateau Limit Tri-Histogram Equalization Algorithm (ATri-PLTri-HEA) is introduced for preprocessing, aiming to improve the quality of the input MRI images and eliminate noise. Then the preprocessed MRI images are segmented using the Masked-attention Mask Transformer (Mask2Former) technique, extracts features through the Fast Point Transformer (FPT), and classified with the Dual-Channel Convolutional Neural Network with Attention-Pooling (DCCNNet-AP) method, which is optimized using the Exponential Distribution Optimizer (EDO). The proposed FaPDCACNNet method is implemented in Python. This approach achieves 98.5% accuracy on three datasets, outperforming existing methods by enhancing feature selection and multi-class classification, making it a trustworthy and precise diagnostic tool for brain tumors.