Scholarlink

<p>The COVID-19 pandemic has spread rapidly across the globe, presenting significant public health challenges. Biomedical imaging techniques, particularly computed tomography (CT), are vital for detecting and monitoring diseases. Accurately segmenting pneumonia lesions in CT scans is essential for diagnosing COVID-19 and assessing the severity of the disease. However, low-contrast infected regions pose a major challenge for automated segmentation methods. In this paper, we present an accessible deep learning framework for the automatic segmentation of COVID-19-infected regions. This framework integrates Contrast-Limited Adaptive Histogram Equalization (CLAHE) preprocessing with an Attention U-Net model trained using a hybrid Dice-Tversky loss. It is supported by extensive data augmentation techniques to improve generalization. We evaluated our approach on a publicly available COVID-19 CT dataset using 5-fold cross-validation. Our results achieved a Dice score of 0.83, an Intersection over Union (IoU) of 0.71, and an accuracy of 99.74%. To enhance the interpretability of our deep learning model, we applied Explainable Artificial Intelligence (XAI) techniques, such as Gradient-weighted Class Activation Mapping (Grad-CAM). These results demonstrate the effectiveness of our proposed framework and highlight its potential as a practical tool for medical imaging applications.</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

Automated segmentation of COVID-19 lesions in CT scans using attention U-net with hybrid loss functions

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年1月11日

查看全文
  • Samy Bakheet,
  • Rehab Youssef,
  • Mahmoud H. Mofaddel,
  • Moatamad Hassan,
  • Asma Alshehri

摘要

The COVID-19 pandemic has spread rapidly across the globe, presenting significant public health challenges. Biomedical imaging techniques, particularly computed tomography (CT), are vital for detecting and monitoring diseases. Accurately segmenting pneumonia lesions in CT scans is essential for diagnosing COVID-19 and assessing the severity of the disease. However, low-contrast infected regions pose a major challenge for automated segmentation methods. In this paper, we present an accessible deep learning framework for the automatic segmentation of COVID-19-infected regions. This framework integrates Contrast-Limited Adaptive Histogram Equalization (CLAHE) preprocessing with an Attention U-Net model trained using a hybrid Dice-Tversky loss. It is supported by extensive data augmentation techniques to improve generalization. We evaluated our approach on a publicly available COVID-19 CT dataset using 5-fold cross-validation. Our results achieved a Dice score of 0.83, an Intersection over Union (IoU) of 0.71, and an accuracy of 99.74%. To enhance the interpretability of our deep learning model, we applied Explainable Artificial Intelligence (XAI) techniques, such as Gradient-weighted Class Activation Mapping (Grad-CAM). These results demonstrate the effectiveness of our proposed framework and highlight its potential as a practical tool for medical imaging applications.