Diagnosing COVID-19 disease using an efficient CAD system.

Area of Science:

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

Background:

• The rapid spread and mutation of COVID-19 necessitate accurate and efficient diagnostic tools.
• Computer-aided diagnosis (CAD) systems offer potential for improving diagnostic speed and accuracy.

Purpose of the Study:

• To propose a novel computer-aided diagnosis (CAD) system, COV-CAD, for the accurate detection of COVID-19.
• To enhance diagnostic capabilities through a combination of feature extraction, classification, and image retrieval.

Main Methods:

• Development of a feature extractor using a modified AlexNet Convolutional Neural Network (CNN) with LeakyReLU activation and a reduced fully connected layer.
• Implementation of a classification method employing majority voting on outputs from content-based image retrieval (CBIR), Support Vector Machine (SVM), K-Nearest Neighbors (KNN), and Random Forest algorithms.
• Integration of a CBIR system to retrieve similar patient cases for further analysis.

Main Results:

• The COV-CAD system achieved high diagnostic accuracy: 93.20% for CT scans and 99.38% for X-ray images.
• The modified AlexNet CNN demonstrated improved efficiency by reducing parameters and training time.
• The majority voting classification approach enhanced overall diagnostic performance.

Conclusions:

• The proposed COV-CAD system provides an efficient and accurate method for COVID-19 diagnosis using medical imaging.
• The integration of advanced AI techniques, including CNNs and CBIR, shows significant promise in combating infectious disease outbreaks.