Explainable Deep Learning for Pulmonary Disease and Coronavirus COVID-19 Detection from X-rays

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Infectious Diseases

Background:

• Coronavirus disease (COVID-19) is a novel infectious disease causing respiratory illness, often presenting with pneumonia.
• Current diagnostic methods for COVID-19 involve sputum or blood tests with results available in hours to days.
• Biomedical imaging, specifically X-rays, can reveal signs of pneumonia associated with COVID-19.

Purpose of the Study:

• To develop a fully automatic and accelerated diagnostic tool for COVID-19 detection.
• To leverage deep learning techniques for analyzing chest X-rays.
• To improve the speed and efficiency of COVID-19 diagnosis.

Main Methods:

• A three-phase deep learning approach was proposed for chest X-ray analysis.
• Phase 1: Pneumonia detection in X-rays.
• Phase 2: Differentiating COVID-19 from other pneumonias.
• Phase 3: Localizing COVID-19 indicative areas within X-rays.

Main Results:

• The proposed deep learning model demonstrated high effectiveness on a dataset of 6,523 chest X-rays.
• Average detection time for COVID-19 was approximately 2.5 seconds.
• The model achieved an average accuracy of 0.97 for COVID-19 detection.

Conclusions:

• The deep learning approach offers a fast and accurate method for COVID-19 detection using chest X-rays.
• This automated system has the potential to significantly expedite the diagnostic process.
• The study validates the efficacy of AI in identifying and localizing COVID-19 indicators in medical imaging.