Scholarlink

<p>Lung diseases, including COVID-19 and pneumonia, are a major global health challenge, necessitating accurate and timely diagnosis. Traditional centralized machine learning approaches for medical image analysis often raise significant privacy concerns due to the sensitive nature of healthcare data. This paper proposes a novel federated learning (FL)-based framework integrated with a custom convolutional neural network (CNN) for the classification of lung diseases using chest X-ray (CXR) images and lung masks. The framework employs a two-stage classification process: an initial binary classification to detect abnormalities and a multiclass classification to identify specific lung diseases such as COVID-19 and viral pneumonia. The work preserves data privacy by decentralizing training and eliminating the need to share sensitive patient data. Experiments conducted on the COVID-19 Radiography Dataset, involving non-IID data distribution across simulated federated environments with 3, 5, and 7 clients, demonstrate competitive performance metrics with an overall accuracy of 93.66%, comparable to state-of-the-art centralized models. This research underscores the potential role of FL in maintaining privacy and enhancing diagnostic accuracy within real-world medical contexts. Moreover, the study utilizes explainable AI methods, particularly Grad-CAM, to interpret and comprehend the results. Furthermore, the integration of AI in healthcare underscores its transformative role in addressing societal challenges and promoting equitable access to advanced diagnostic tools while upholding ethical considerations.</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

Convolutional neural network (CNN) and federated learning-based approach for lung disease detection

  • 期刊论文

  • 发表日期：

    2025年8月25日

查看全文
  • Mithun Karmakar,
  • Ashlesha Hota,
  • Amitava Nag

摘要

Lung diseases, including COVID-19 and pneumonia, are a major global health challenge, necessitating accurate and timely diagnosis. Traditional centralized machine learning approaches for medical image analysis often raise significant privacy concerns due to the sensitive nature of healthcare data. This paper proposes a novel federated learning (FL)-based framework integrated with a custom convolutional neural network (CNN) for the classification of lung diseases using chest X-ray (CXR) images and lung masks. The framework employs a two-stage classification process: an initial binary classification to detect abnormalities and a multiclass classification to identify specific lung diseases such as COVID-19 and viral pneumonia. The work preserves data privacy by decentralizing training and eliminating the need to share sensitive patient data. Experiments conducted on the COVID-19 Radiography Dataset, involving non-IID data distribution across simulated federated environments with 3, 5, and 7 clients, demonstrate competitive performance metrics with an overall accuracy of 93.66%, comparable to state-of-the-art centralized models. This research underscores the potential role of FL in maintaining privacy and enhancing diagnostic accuracy within real-world medical contexts. Moreover, the study utilizes explainable AI methods, particularly Grad-CAM, to interpret and comprehend the results. Furthermore, the integration of AI in healthcare underscores its transformative role in addressing societal challenges and promoting equitable access to advanced diagnostic tools while upholding ethical considerations.