Scholarlink

Glioblastoma is an aggressive cancer that can occur in the brain or spinal cord. Glioma is a dangerous brain tumour that affects many individuals. Tumour-treating fields (TTF) involve applying adhesive pads to your scalp. Palliative care can be used while undergoing other treatments, such as radiation therapy or chemo-by-needle. The risk of side effects may not be known from clinical trials. Glioma segmentation is a frequent method used by physicians. Using this procedure, the appearance of glioma is examined, analysed, and diagnosed. Gliomas are often segmented using deep learning. This study solves the problem by segmenting gliomas using an improved convolutional neural network (CNN) model. Although 3D full CNNs have the benefit of gathering 3D spatial data, they need more memory than 2D full CNNs. The DenseBlock encoder, decoder, and attention units are proposed in the model to segment gliomas. Using multisequence glioma imaging data lowers the 3Dness of the UNet model on the BraTS17 validation set, but it worked well. BraTS1 is about as sensitive as the top segmentation model. It also correctly identified gliomas. The dice, Hausdorff distance, sensitivity, and specificity of the segmented tumour are all equal to 0.8492. The method works for both high-grade and low-grade gliomas. The proposed strategy outperforms competing techniques in testing.

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

Detection of Gliomas Using a Novel Attention UNet-Based Deep Learning Model

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Yasmin Ammar Adi,
  • Mohammed Khaleel Jameel,
  • Mohammed Ahmed Mustafa,
  • Ekhlas A. K. Kanani,
  • Hawraa Ali Sabah,
  • Rajaa Jasim Mohammed,
  • Heba A. Abd-Alsalam Alsalame,
  • Rajit Nair

摘要

Glioblastoma is an aggressive cancer that can occur in the brain or spinal cord. Glioma is a dangerous brain tumour that affects many individuals. Tumour-treating fields (TTF) involve applying adhesive pads to your scalp. Palliative care can be used while undergoing other treatments, such as radiation therapy or chemo-by-needle. The risk of side effects may not be known from clinical trials. Glioma segmentation is a frequent method used by physicians. Using this procedure, the appearance of glioma is examined, analysed, and diagnosed. Gliomas are often segmented using deep learning. This study solves the problem by segmenting gliomas using an improved convolutional neural network (CNN) model. Although 3D full CNNs have the benefit of gathering 3D spatial data, they need more memory than 2D full CNNs. The DenseBlock encoder, decoder, and attention units are proposed in the model to segment gliomas. Using multisequence glioma imaging data lowers the 3Dness of the UNet model on the BraTS17 validation set, but it worked well. BraTS1 is about as sensitive as the top segmentation model. It also correctly identified gliomas. The dice, Hausdorff distance, sensitivity, and specificity of the segmented tumour are all equal to 0.8492. The method works for both high-grade and low-grade gliomas. The proposed strategy outperforms competing techniques in testing.