Prediction of COVID-19 Using Genetic Deep Learning Convolutional Neural Network (GDCNN)

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Pneumonia Diagnosis

Background:

• Coronavirus disease (COVID-19) rapid spread causes severe pneumonia, impacting healthcare systems.
• Early diagnosis of COVID-19 pneumonia is crucial for effective treatment and reducing healthcare burden.
• Chest X-ray (CXR) is a widely accessible, rapid, and cost-effective imaging modality for pneumonia diagnosis.

Purpose of the Study:

• To develop and evaluate a deep learning model for accurate COVID-19 pneumonia detection using CXR images.
• To compare the performance of the proposed model against established transfer learning techniques.
• To provide a reliable tool for differentiating COVID-19 pneumonia from normal lung conditions.

Main Methods:

• Utilized a Genetic Deep Learning Convolutional Neural Network (GDCNN) trained from scratch.
• Employed a dataset of over 5000 CXR images for training and classification.
• Compared GDCNN performance against models like ReseNet18, ReseNet50, Squeezenet, DenseNet-121, and VGG16.

Main Results:

• Achieved a classification accuracy of 98.84% for COVID-19 prediction.
• Demonstrated high performance metrics: 93% precision, 100% sensitivity, and 97.0% specificity.
• The GDCNN model outperformed existing transfer learning techniques in identifying COVID-19 pneumonia.

Conclusions:

• The novel GDCNN model offers a highly accurate and efficient method for COVID-19 pneumonia detection via CXR.
• The proposed approach shows superior performance in an unbalanced dataset environment.
• This research contributes a valuable tool for early and precise COVID-19 diagnosis, alleviating healthcare system pressure.