Scholarlink

<p>Brain tumor detection using magnetic resonance imaging (MRI) is vital for early diagnosis, but conventional methods often struggle with accuracy and real-time performance. This work introduces a novel cloud-supported Internet of Medical Things (IoMT) framework for efficient brain tumor classification. The proposed model, named Zeiler–Fergus Memory-Boosted Dilated Progressive Feedback Residual Attention Network with Dung Beetle Optimizer (ZFMBDPFRANet-DBO) integrates a Memory-Boosted Twin Guidance Filtering (MBTGF) mechanism to enhance MRI quality, followed by feature extraction using a Zeiler and Fergus Residual Network (ZFNet-DRN). A Dilated Progressive Feedback Residual Attention Network (DCPFRANet) is employed for classification, and optimization is carried out using the Dung Beetle Optimizer (DBO). Testing out BraTS 2020 data set has an exceptional result of accuracy: 99.97% on normal, 99.84 percent on glioma, 99.60 percent on pituitary, and 99.87% on meningioma. The sensitivity and specificity are above 99%, the system has minimal delay (10.34&#xa0;ms) and high throughput (1250 kbps), which is a key feature applicable in remote e-health cases. This architecture provides a scalable, performance-oriented framework to permit real-time diagnosis of brain tumor in healthcare systems equipped with IoMT.</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

Cloud IoMT-Enabled Brain Tumor Detection Using Optimized Dilated Residual Attention Network

  • 期刊论文

  • 发表日期：

    2025年8月5日

查看全文
  • Anuradha Taluja,
  • Harish Kumar,
  • D. Bhuvana Suganthi,
  • Elangovan Muniyandy

摘要

Brain tumor detection using magnetic resonance imaging (MRI) is vital for early diagnosis, but conventional methods often struggle with accuracy and real-time performance. This work introduces a novel cloud-supported Internet of Medical Things (IoMT) framework for efficient brain tumor classification. The proposed model, named Zeiler–Fergus Memory-Boosted Dilated Progressive Feedback Residual Attention Network with Dung Beetle Optimizer (ZFMBDPFRANet-DBO) integrates a Memory-Boosted Twin Guidance Filtering (MBTGF) mechanism to enhance MRI quality, followed by feature extraction using a Zeiler and Fergus Residual Network (ZFNet-DRN). A Dilated Progressive Feedback Residual Attention Network (DCPFRANet) is employed for classification, and optimization is carried out using the Dung Beetle Optimizer (DBO). Testing out BraTS 2020 data set has an exceptional result of accuracy: 99.97% on normal, 99.84 percent on glioma, 99.60 percent on pituitary, and 99.87% on meningioma. The sensitivity and specificity are above 99%, the system has minimal delay (10.34 ms) and high throughput (1250 kbps), which is a key feature applicable in remote e-health cases. This architecture provides a scalable, performance-oriented framework to permit real-time diagnosis of brain tumor in healthcare systems equipped with IoMT.