A Seven-Layer Convolutional Neural Network for Chest CT-Based COVID-19 Diagnosis Using Stochastic Pooling

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Infectious Diseases

Background:

• The COVID-19 pandemic has caused significant mortality globally.
• Chest computed tomography (CT) is a crucial tool for diagnosing COVID-19.
• Accurate and efficient diagnostic methods are essential for pandemic management.

Purpose of the Study:

• To develop and evaluate a novel seven-layer convolutional neural network (7L-CNN-CD) for automated COVID-19 diagnosis using chest CT images.
• To enhance the performance of the diagnostic model through innovative data augmentation and pooling techniques.

Main Methods:

• A seven-layer convolutional neural network (7L-CNN-CD) architecture was designed for COVID-19 detection.
• A 14-way data augmentation strategy was implemented to expand the training dataset.
• Stochastic pooling was utilized as an alternative to traditional pooling methods.

Main Results:

• The 7L-CNN-CD model demonstrated a high sensitivity of 94.44±0.73% and specificity of 93.63±1.60%.
• The overall accuracy achieved by the model was 94.03±0.80% across 10 runs of 10-fold cross-validation.
• The proposed data augmentation and stochastic pooling methods significantly contributed to the model's effectiveness.

Conclusions:

• The 7L-CNN-CD model is a highly effective tool for diagnosing COVID-19 from chest CT images.
• The proposed model exhibits superior performance compared to several state-of-the-art diagnostic algorithms.
• Data augmentation and stochastic pooling are validated as beneficial techniques for improving deep learning-based medical image analysis.