Scholarlink

<p>The paper presents an efficient brain tumor segmentation and classification. Initially, the Pixel density based trimmed median filter (PDTMF) is utilized to pre-process the image, which enhances its quality. Further, an Improved Attention based U-Net (IAU-Net) model is utilized to achieve efficient segmentation. It also reduces the issue of insufficient small scale brain tumor and enhances the segmentation accuracy. The Deep Residual Auto-encoder (deep RAE) is used to extract the feature automatically from the segmented tumor region. Later, on the extracted features, the dung beetle optimizer-based hybrid gated recurrent unit (DBO-HGRU) is utilized to classify the brain tumor. Dung beetle optimizer (DBO) tunes the hyper-parameters, which enhance the overall performance. In this research work, the Brats 2020 dataset is utilized, and it attained 99.3% accuracy, which is higher than other related techniques. The precision, recall, specificity, and F1-score of the proposed technique are obtained to be 98.35%, 98.73%, 98.3%, and 99.73%.</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

DBO-HGRU-enhanced brain tumor segmentation and classification technique using auto-encoder

  • 期刊论文

  • 发表日期：

    2026年1月27日

查看全文
  • Amit Kumar Gupta,
  • Harsh Khatter,
  • Ruchi Rani,
  • Anjali Jain,
  • Mangal Sain

摘要

The paper presents an efficient brain tumor segmentation and classification. Initially, the Pixel density based trimmed median filter (PDTMF) is utilized to pre-process the image, which enhances its quality. Further, an Improved Attention based U-Net (IAU-Net) model is utilized to achieve efficient segmentation. It also reduces the issue of insufficient small scale brain tumor and enhances the segmentation accuracy. The Deep Residual Auto-encoder (deep RAE) is used to extract the feature automatically from the segmented tumor region. Later, on the extracted features, the dung beetle optimizer-based hybrid gated recurrent unit (DBO-HGRU) is utilized to classify the brain tumor. Dung beetle optimizer (DBO) tunes the hyper-parameters, which enhance the overall performance. In this research work, the Brats 2020 dataset is utilized, and it attained 99.3% accuracy, which is higher than other related techniques. The precision, recall, specificity, and F1-score of the proposed technique are obtained to be 98.35%, 98.73%, 98.3%, and 99.73%.