Scholarlink

<p>Stroke detection and classification from computed tomography (CT) remains a critical and challenging task in medical imaging due to the complexity of lesion patterns, noise variations and unbalanced datasets. In this study, we propose a novel hybrid deep learning model, StrokeFuse-AttnNet, which integrates both global (ResNet50) and local (DenseNet121) convolutional feature extractors with a self-attention mechanism to improve spatial focus and semantic interpretability. A hierarchical feature fusion strategy concatenates multi-scale features, which are then processed by a self-attention module to highlight key stroke regions and reduce irrelevant activations. We use data augmentation and SMOTE on training samples to address imbalance and improve generalization. The proposed model was evaluated on both publicly and privately available brain CT datasets. StrokeFuse-AttnNet achieved an accuracy of 98.27% and an AUC of 0.983 on the public dataset and an accuracy of 96.04% and an AUC of 0.9501 on the private dataset. The results show that the model has higher accuracy, reliability and generalization than existing and baseline methods. The proposed model is lightweight, with only 32 million parameters and can be used in real-time clinical diagnostic processing systems that require 40 GFLOPs. The model has the potential to support radiologists in the efficient and rapid diagnosis of strokes on non-contrast CT images.</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

StrokeFuse-AttnNet: a hybrid feature fusion and self-attention model for stroke detection using neuroimages

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年3月31日

查看全文
  • Muhammad Asim Saleem,
  • Ashir Javeed,
  • Wasan Akarathanawat,
  • Aurauma Chutinet,
  • Nijasri Charnnarong Suwanwela,
  • Pasu Kaewplung,
  • Surachai Chaitusaney,
  • Watit Benjapolakul

摘要

Stroke detection and classification from computed tomography (CT) remains a critical and challenging task in medical imaging due to the complexity of lesion patterns, noise variations and unbalanced datasets. In this study, we propose a novel hybrid deep learning model, StrokeFuse-AttnNet, which integrates both global (ResNet50) and local (DenseNet121) convolutional feature extractors with a self-attention mechanism to improve spatial focus and semantic interpretability. A hierarchical feature fusion strategy concatenates multi-scale features, which are then processed by a self-attention module to highlight key stroke regions and reduce irrelevant activations. We use data augmentation and SMOTE on training samples to address imbalance and improve generalization. The proposed model was evaluated on both publicly and privately available brain CT datasets. StrokeFuse-AttnNet achieved an accuracy of 98.27% and an AUC of 0.983 on the public dataset and an accuracy of 96.04% and an AUC of 0.9501 on the private dataset. The results show that the model has higher accuracy, reliability and generalization than existing and baseline methods. The proposed model is lightweight, with only 32 million parameters and can be used in real-time clinical diagnostic processing systems that require 40 GFLOPs. The model has the potential to support radiologists in the efficient and rapid diagnosis of strokes on non-contrast CT images.