Scholarlink

<p>Pneumonia is caused due to various microorganisms and respiratory infections. A timely and accurate diagnosis is required to ensure effective precaution. Currently, Deep Learning (DL) has emerged as a powerful and widely adopted approach for detecting pneumonia using Chest X-Ray (CXR) images. Traditional diagnostic methods, such as manual examination of CXR by radiologists, are subject to human error and variability. Therefore, this paper presents the Attention-Pyramid Shuffling Taylor Network (APSFT-Net) for pneumonia detection. The process begins with the acquisition of images from relevant datasets, and then noises presented in the CXR images are removed by the Wiener filter. Lung lobes are then segmented utilizing Dual-Attention V-Network (DAV-Net). Then, image augmentation is applied, including image erasing, manipulation, and mix techniques. Feature extraction is performed using both Gray Level Run Length Matrix (GLRM) and Dual Tree Local Ternary Pattern (DTLTP), which combines Local Ternary Pattern (LTP) and Dual-Tree Complex Wavelet Transform (DTCWT). Finally, pneumonia detection is done using APSFT-Net that integrates Shuffle Attention Network (SA-Net), Multi-scale Feature Pyramid Network (MFPN) and Taylor series. At a 90% learning set, APSFT-Net achieved an accuracy of 97.190%, True Positive Rate (TPR) of 97.857%, True Negative Rate (TNR) of 96.777%, precision of 96.222%, and F1-score of 97.033% using the CXR Image (Pneumonia) Database.</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

APSFT-Net: Attention-Pyramid Shuffling Taylor Network for Accurate Pneumonia Detection in Chest X-Rays

  • 期刊论文

  • 发表日期：

    2026年2月23日

查看全文
  • G. Nirmalapriya,
  • Belsam Jeba Ananth. M,
  • D. Anu Disney,
  • Santhoshini Sahu

摘要

Pneumonia is caused due to various microorganisms and respiratory infections. A timely and accurate diagnosis is required to ensure effective precaution. Currently, Deep Learning (DL) has emerged as a powerful and widely adopted approach for detecting pneumonia using Chest X-Ray (CXR) images. Traditional diagnostic methods, such as manual examination of CXR by radiologists, are subject to human error and variability. Therefore, this paper presents the Attention-Pyramid Shuffling Taylor Network (APSFT-Net) for pneumonia detection. The process begins with the acquisition of images from relevant datasets, and then noises presented in the CXR images are removed by the Wiener filter. Lung lobes are then segmented utilizing Dual-Attention V-Network (DAV-Net). Then, image augmentation is applied, including image erasing, manipulation, and mix techniques. Feature extraction is performed using both Gray Level Run Length Matrix (GLRM) and Dual Tree Local Ternary Pattern (DTLTP), which combines Local Ternary Pattern (LTP) and Dual-Tree Complex Wavelet Transform (DTCWT). Finally, pneumonia detection is done using APSFT-Net that integrates Shuffle Attention Network (SA-Net), Multi-scale Feature Pyramid Network (MFPN) and Taylor series. At a 90% learning set, APSFT-Net achieved an accuracy of 97.190%, True Positive Rate (TPR) of 97.857%, True Negative Rate (TNR) of 96.777%, precision of 96.222%, and F1-score of 97.033% using the CXR Image (Pneumonia) Database.