Classification of COVID-19 from tuberculosis and pneumonia using deep learning techniques

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Deep Learning

Background:

• Accurate prediction of lung diseases like tuberculosis, pneumonia, and COVID-19 is challenging due to imbalanced medical image datasets.
• Conventional machine learning methods struggle with limited data, particularly for diseases like COVID-19.

Purpose of the Study:

• To develop an effective deep learning model for accurate and timely classification of multiple lung diseases.
• To address data imbalance issues in medical imaging datasets for improved diagnostic performance.

Main Methods:

• Utilized deep learning, specifically convolutional neural networks, for image analysis.
• Implemented image data augmentation and Synthetic Minority Oversampling Technique (SMOTE) for class balancing.
• Developed a multi-level classification system for differentiating tuberculosis, pneumonia, and COVID-19.

Main Results:

• Achieved high classification accuracy: 97.4% for tuberculosis and pneumonia, and 88% for bacterial, viral, and COVID-19 classifications.
• Demonstrated an approximate 8-10% improvement in classification accuracy compared to existing methods.
• The proposed system shows scalability with growing medical data and faster classification.

Conclusions:

• Combined data augmentation and class balancing techniques significantly improve lung disease classification accuracy.
• The developed multi-level classification model offers a scalable and accurate solution for diagnosing lung diseases and their subtypes.
• Deep learning with advanced data handling techniques shows great promise for medical diagnostics.