Deep Learning-Aided Automated Pneumonia Detection and Classification Using CXR Scans

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Pneumonia Diagnosis

Background:

• The COVID-19 pandemic strained healthcare systems globally.
• Reverse transcription-polymerase chain reaction (RT-PCR) has limitations in COVID-19 detection accuracy (70%).
• COVID-19 commonly causes pneumonia, making chest X-rays a potential diagnostic tool.

Purpose of the Study:

• To develop and evaluate an AI model for detecting COVID-19-induced pneumonia from chest X-rays.
• To differentiate COVID-19 pneumonia from regular pneumonia using deep learning.
• To improve diagnostic accuracy beyond traditional methods.

Main Methods:

• Utilized Convolutional Neural Network (CNN) and deep learning techniques.
• Employed transfer learning with fine-tuning using Xception, VGG16, and VGG19 models.
• Classified chest X-rays into three categories: COVID-19 pneumonia, regular pneumonia, and normal.

Main Results:

• Achieved a 98% accuracy in detecting COVID-19-induced pneumonia.
• Demonstrated high performance across various metrics including precision, recall, and F1 score.
• Successfully distinguished between COVID-19 pneumonia and common pneumonia, addressing diagnostic overlap.

Conclusions:

• The proposed AI model shows significant promise for accurate COVID-19 pneumonia detection from chest X-rays.
• This approach can aid in differentiating lung infections, offering a more reliable diagnostic standard.
• The AI model can help mitigate diagnostic complexities and improve patient outcomes.