Scholarlink

<p>Pneumonia is a life-threatening disease that has affected millions of people worldwide. Advancements in technology have improved the speedy detection of pneumonia, which supports early diagnosis and treatment. However, an accurate pneumonia diagnosis is crucial, and unsatisfactory treatment causes serious consequences for patients and may become fatal. Deep learning techniques have contributed to supporting the medical experts in diagnosing pneumonia. Even though a vast number of methodologies are adopted for pneumonia detection, the inefficiency of the existing model affects their practical applicability. Hence, this research develops a Lightweight Double Attention-based Deep Bidirectional Gated Recurrent Long Short-Term Memory (LD-DBGRTM) model to detect pneumonia using CT images. The proposed research incorporates hybrid statistical features and saliency map features for forming the feature vector, which forms the input to the detection module. The evaluation of the model illustrates that the proposed LD-DBGRTM model achieved superior performance with the highest accuracy of 99.28%, a negative predictive value (NPV) of 0.99, and a positive predictive value (PPV) of 0.99 during the training percentage analysis.</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

LD-DBGRTM: Vision Transformer Enabled Lightweight Attention-Based Deep Learning Model for Pneumonia Detection

  • 期刊论文

  • 发表日期：

    2025年10月19日

查看全文
  • Sanchi Kaushik,
  • Ruchi Verma

摘要

Pneumonia is a life-threatening disease that has affected millions of people worldwide. Advancements in technology have improved the speedy detection of pneumonia, which supports early diagnosis and treatment. However, an accurate pneumonia diagnosis is crucial, and unsatisfactory treatment causes serious consequences for patients and may become fatal. Deep learning techniques have contributed to supporting the medical experts in diagnosing pneumonia. Even though a vast number of methodologies are adopted for pneumonia detection, the inefficiency of the existing model affects their practical applicability. Hence, this research develops a Lightweight Double Attention-based Deep Bidirectional Gated Recurrent Long Short-Term Memory (LD-DBGRTM) model to detect pneumonia using CT images. The proposed research incorporates hybrid statistical features and saliency map features for forming the feature vector, which forms the input to the detection module. The evaluation of the model illustrates that the proposed LD-DBGRTM model achieved superior performance with the highest accuracy of 99.28%, a negative predictive value (NPV) of 0.99, and a positive predictive value (PPV) of 0.99 during the training percentage analysis.