Scholarlink

The importance of rapid and accurate diagnostic tools for pneumonia is still very important, given that pneumonia is leading cause of global morbidity and mortality. Therefore, a stacking deep learning ensemble system built on top of VGG16 convolutional neural networks is developed for podiatric chest X-ray pneumonia classification. The inputs of the XGBoost metal earner are deep features extracted by VGG16 from pre-processed images (resized, normalized, and augmented). Finally, on a collected set of 5,863 images, the stacking model reaches 99.68% accuracy and 0.998 AUC-ROC better than individual CNN baselines. It is comparatively analysed to some of the recent literature and is seen to provide better diagnostic precision. The results prove the feasibility of employing AI based automated detection tools in early pneumonia. Further work in real time clinical deployment, integration of multimodal medical data, and interpretability as well computational efficiency will be performed.

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

CNN Based Pneumonia Detection Model with Ensemble Methods Using Chest X-Ray Images

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Akshit Jain,
  • Sandeep Singh Kang,
  • Neha Dutta

摘要

The importance of rapid and accurate diagnostic tools for pneumonia is still very important, given that pneumonia is leading cause of global morbidity and mortality. Therefore, a stacking deep learning ensemble system built on top of VGG16 convolutional neural networks is developed for podiatric chest X-ray pneumonia classification. The inputs of the XGBoost metal earner are deep features extracted by VGG16 from pre-processed images (resized, normalized, and augmented). Finally, on a collected set of 5,863 images, the stacking model reaches 99.68% accuracy and 0.998 AUC-ROC better than individual CNN baselines. It is comparatively analysed to some of the recent literature and is seen to provide better diagnostic precision. The results prove the feasibility of employing AI based automated detection tools in early pneumonia. Further work in real time clinical deployment, integration of multimodal medical data, and interpretability as well computational efficiency will be performed.