Scholarlink

<p>A hemorrhage that occurs inside the skull is known as an intracranial hemorrhage. Due to the high fatality rate and the potential for long-term impairment, intracranial bleeding requires rapid medical intervention. Because intracranial hemorrhage can come in a broad variety of sizes, severity levels, and morphologies, it can be difficult to accurately identify the condition. Among the many imaging modalities that are available, computed tomography (CT) imaging is considered to be among the most significant modalities for the purpose of analyzing brain images. The proposed work focuses primarily on a novel deep attention recurrent residual U-Net model (DAR2U-Net) for the purpose of segmenting CT images of intracranial hemorrhage. Intersection over union and binary focal loss are the performance evaluation matrices that are utilized in order to conduct an analysis of the training and validation loss. Comparisons are made between the proposed model and the U-Net and Attention U-Net models. The experimental result that was able to acquire the greatest possible score for the suggested model’s intersection over union was 0.5810.</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

Intracranial Hemorrhages Segmentation Using Deep Attention Recurrent Residual U-Net (DAR2U-Net) Model

  • 期刊论文

  • 发表日期：

    2026年1月19日

查看全文
  • Santwana S. Gudadhe,
  • Anuradha D. Thakare

摘要

A hemorrhage that occurs inside the skull is known as an intracranial hemorrhage. Due to the high fatality rate and the potential for long-term impairment, intracranial bleeding requires rapid medical intervention. Because intracranial hemorrhage can come in a broad variety of sizes, severity levels, and morphologies, it can be difficult to accurately identify the condition. Among the many imaging modalities that are available, computed tomography (CT) imaging is considered to be among the most significant modalities for the purpose of analyzing brain images. The proposed work focuses primarily on a novel deep attention recurrent residual U-Net model (DAR2U-Net) for the purpose of segmenting CT images of intracranial hemorrhage. Intersection over union and binary focal loss are the performance evaluation matrices that are utilized in order to conduct an analysis of the training and validation loss. Comparisons are made between the proposed model and the U-Net and Attention U-Net models. The experimental result that was able to acquire the greatest possible score for the suggested model’s intersection over union was 0.5810.