Deep Content Information Retrieval for COVID-19 Detection from Chromatic CT Scans

Area of Science:

• Medical Imaging Analysis
• Artificial Intelligence in Healthcare
• Infectious Disease Diagnostics

Background:

• Computed tomography (CT) scans are crucial for diagnosing pneumonia, a common complication of COVID-19.
• Accurate and rapid detection of COVID-19 related pneumonia is essential for patient management and prognostication.
• The role of chromatic information in CT scans for disease detection requires further investigation.

Purpose of the Study:

• To investigate the utility of chromatic information in CT scans for COVID-19 detection.
• To confirm the effectiveness of artificial intelligence (AI) findings in assisting COVID-19 diagnosis.
• To propose and evaluate a novel AI model for classifying COVID-19 pneumonia.

Main Methods:

• Development of a freezing-based convolutional neural network (CNN) learning approach.
• Application of morphological transformations to CT images for feature extraction.
• Classification of COVID-19 cohorts using the proposed CNN architecture on existing CT image datasets.

Main Results:

• The proposed CNN model demonstrated a powerful ability to recognize lesions in CT images.
• The AI model achieved a diagnostic accuracy of 87.56% for COVID-19 detection.
• An improvement of 3% in accuracy was observed compared to the baseline method.

Conclusions:

• Chromatic information in CT scans, when processed by AI, significantly aids in COVID-19 detection.
• The developed CNN architecture effectively reflects viral proliferation in infected patients.
• The AI-assisted approach offers enhanced accuracy and speed for pneumonia disease monitoring.