Automated COVID-19 detection from X-ray and CT images with stacked ensemble convolutional neural network

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Computer-Aided Diagnosis

Background:

• The COVID-19 pandemic necessitates rapid and accurate diagnostic tools.
• Distinguishing COVID-19 from other pneumonias using radiological images presents challenges.
• Automated screening of chest X-rays and CT scans is crucial for timely patient management.

Purpose of the Study:

• To develop and evaluate a stacked convolutional neural network (CNN) model for automated COVID-19 diagnosis.
• To improve the accuracy and reliability of detecting COVID-19 from chest X-ray and CT images.
• To compare the performance of the proposed stacked CNN model against existing methods.

Main Methods:

• A novel stacked CNN architecture was designed, integrating sub-models from VGG19 and Xception.
• A softmax classifier was employed to combine the outputs of the sub-models.
• Datasets were created using publicly available X-ray repositories and collected CT images.
• The model was trained and validated for multi-class (X-ray) and binary (CT) classification tasks.

Main Results:

• The stacked CNN model achieved 97.62% sensitivity for multi-class classification of X-ray images (COVID-19, Normal, Pneumonia).
• The model demonstrated 98.31% sensitivity for binary classification of CT images (COVID-19 vs. no-Finding).
• The proposed approach exhibited superior performance compared to existing methods for COVID-19 detection from X-ray images.

Conclusions:

• The developed stacked CNN model offers a highly sensitive and accurate method for automated COVID-19 detection.
• This AI-driven approach can aid in the rapid screening and diagnosis of COVID-19 using radiological imaging.
• The model's superior performance highlights its potential clinical utility in pandemic situations.