Scholarlink

Pneumonia is still a serious worldwide health issue, particularly in areas with lack of resources, where getting a prompt and accurate diagnosis is difficult. Early detection is crucial to minimize mortality rates and facilitate proper treatment. To address this problem, deep learning based approaches are applied to detect pneumonia. The system uses data pre-processing methods like image normalization and classified as normal or positive for pneumonia by using VGG-19 convolutional neural network (CNN). Furthermore, Grad-CAM is used to identify the areas in the chest X-rays causing Pneumonia. Grad-CAM highlights the specific areas in the lungs images that led the system to make a clear decision about the results of the diagnosis process which was made. This visual highlighting increases the clinical trust and usability of the system. Benchmark open-source datasets are used to train and test the system. With the integration of high-accuracy classification model and explainable AI, the proposed system assist radiologists and healthcare professionals with more precise diagnoses. This will enhance patient care, trust and decreases diagnostic errors in medical environments.

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

Pneumonia Detection VGG- 19 and Grad-CAM for Explainable Deep Learning Diagnosis

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Mallannagari Sunitha,
  • Vishwanala Udith,
  • Mandadi Vishnu Vardhan Reddy,
  • Marya Coka

摘要

Pneumonia is still a serious worldwide health issue, particularly in areas with lack of resources, where getting a prompt and accurate diagnosis is difficult. Early detection is crucial to minimize mortality rates and facilitate proper treatment. To address this problem, deep learning based approaches are applied to detect pneumonia. The system uses data pre-processing methods like image normalization and classified as normal or positive for pneumonia by using VGG-19 convolutional neural network (CNN). Furthermore, Grad-CAM is used to identify the areas in the chest X-rays causing Pneumonia. Grad-CAM highlights the specific areas in the lungs images that led the system to make a clear decision about the results of the diagnosis process which was made. This visual highlighting increases the clinical trust and usability of the system. Benchmark open-source datasets are used to train and test the system. With the integration of high-accuracy classification model and explainable AI, the proposed system assist radiologists and healthcare professionals with more precise diagnoses. This will enhance patient care, trust and decreases diagnostic errors in medical environments.