Scholarlink

Pneumonia is a serious lung infection that, if not treated promptly, can result in major health issues. Traditionally, chest X-rays (CXR) have been the primary means of diagnosing pneumonia; nevertheless, their precision continues to rely on the skill of radiologists, which can lead to delays and inconsistent results. In this study, we present a deep neural network algorithm-based automated pneumonia detection method using X-ray images. The Chest X-Ray Images (Pneumonia) collection contains 5,850 labeled images. To enhance feature extraction from such images, Contrast Limited Adaptive Histogram Equalization (CLAHE) was used for image preparation. We employ convolutional neural networks (CNNs) to classify X-ray images into lung segments affected by pneumonia and those that are clean utilizing the InceptionV3 and ResNet50V2 architectures. Transfer learning does much to improve this type of training for efficiency. Preliminary results using ResNet50V2 yielded a model that showed 99.00% training accuracy and 92.71% validation accuracy. This approach should avail radiologists of a robust tool for computerized analysis by augmenting the efficacy in pneumonia detection. More assessments and studies will be done to determine the overall efficacy of the model.

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 Detection of Pneumonia Using Chest X-Ray Images with Deep Convolutional Neural Networks

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Pulak Islam,
  • Md. Sabbir Hossain,
  • Nabila Ibnat,
  • Lamia Ibnat,
  • Ummay Mariom Sumi,
  • Fatema Jahan Rumi,
  • Sadman Sakib,
  • Md. Saiful Islam

摘要

Pneumonia is a serious lung infection that, if not treated promptly, can result in major health issues. Traditionally, chest X-rays (CXR) have been the primary means of diagnosing pneumonia; nevertheless, their precision continues to rely on the skill of radiologists, which can lead to delays and inconsistent results. In this study, we present a deep neural network algorithm-based automated pneumonia detection method using X-ray images. The Chest X-Ray Images (Pneumonia) collection contains 5,850 labeled images. To enhance feature extraction from such images, Contrast Limited Adaptive Histogram Equalization (CLAHE) was used for image preparation. We employ convolutional neural networks (CNNs) to classify X-ray images into lung segments affected by pneumonia and those that are clean utilizing the InceptionV3 and ResNet50V2 architectures. Transfer learning does much to improve this type of training for efficiency. Preliminary results using ResNet50V2 yielded a model that showed 99.00% training accuracy and 92.71% validation accuracy. This approach should avail radiologists of a robust tool for computerized analysis by augmenting the efficacy in pneumonia detection. More assessments and studies will be done to determine the overall efficacy of the model.