Simultaneous Super-Resolution and Classification of Lung Disease Scans

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Pulmonology

Background:

• Acute lower respiratory infections are a major cause of mortality, especially in developing nations.
• Current diagnostic methods require improvement, particularly in resource-limited environments.
• Artificial intelligence (AI) applications in analyzing chest X-rays and computed tomography (CT) for these infections are underexplored.

Purpose of the Study:

• To develop and evaluate a computer-aided diagnostic (CAD) system for detecting common pulmonary diseases from chest X-ray and CT images.
• To enhance image quality using super-resolution (SR) techniques for improved diagnostic accuracy.
• To leverage deep learning (DL) for both image reconstruction and disease classification.

Main Methods:

• The proposed system integrates super-resolution (SR) techniques with deep learning (DL) models.
• The InceptionResNetv2 model was employed as a feature extractor, coupled with a multi-class support vector machine (MCSVM) classifier.
• Performance was benchmarked against other models like Resnet101 and Inceptionv3, and classifier effectiveness (softmax vs. MCSVM) was assessed on public datasets.

Main Results:

• The developed system achieved a high classification accuracy of 98.028%.
• The combination of SR techniques and the InceptionResNetv2 model demonstrated superior performance.
• The system proved effective in classifying various pulmonary conditions including COVID-19, pneumonia, tuberculosis, lung opacity, and carcinoma.

Conclusions:

• The AI-powered system shows significant potential as a screening tool for lower respiratory disorders.
• It can assist clinicians in interpreting complex chest imaging data, enhancing diagnostic capabilities.
• The system offers valuable diagnostic support, especially in resource-limited healthcare settings.