A multi model ensemble based deep convolution neural network structure for detection of COVID19.

Area of Science:

• Artificial Intelligence
• Medical Imaging
• Computer-Aided Diagnosis

Background:

• The COVID-19 pandemic caused a global health crisis.
• Traditional diagnostic methods face challenges in high-density populations.
• Accurate and rapid detection of COVID-19 is crucial.

Purpose of the Study:

• To develop an automated model for detecting COVID-19 using chest X-ray images.
• To improve diagnostic accuracy and efficiency for radiologists.
• To address limitations in testing availability.

Main Methods:

• Utilized an ensemble of four pre-trained Deep Convolutional Neural Network (DCNN) architectures (VGGNet, GoogleNet, DenseNet, NASNet) for feature extraction.
• Extracted low-level features from chest X-ray images.
• Employed a fully connected layer for classification.

Main Results:

• Achieved 88.98% accuracy for three-class classification on public datasets.
• Reported 98.58% accuracy for binary classification on public datasets.
• Obtained 93.48% accuracy on a private dataset.

Conclusions:

• The proposed multi-model ensemble architecture outperforms single classifiers.
• The DCNN-based approach shows significant potential for COVID-19 detection from chest X-rays.
• The developed model can assist radiologists in diagnosing COVID-19, especially during outbreaks.