Classification of COVID-19 Chest CT Images Based on Ensemble Deep Learning.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Computer Science

Background:

• Novel coronavirus pneumonia (NCP) poses a global health challenge.
• Computed tomography (CT) imaging is crucial for NCP diagnosis.
• Efficient diagnostic tools are needed to manage the pandemic.

Purpose of the Study:

• To develop an assisted diagnosis algorithm for NCP using ensemble deep learning.
• To improve the efficiency and speed of NCP diagnosis for medical personnel.
• To enhance multi-category prediction performance in deep neural networks.

Main Methods:

• An ensemble deep learning algorithm combining Stacked Generalization and VGG16 was developed.
• A cascade classifier was formed using multiple training data subsets.
• The algorithm was validated for classifying NCP, common pneumonia (CP), and normal controls.

Main Results:

• The algorithm achieved 93.57% prediction accuracy for the three categories.
• Sensitivity reached 94.21%, specificity 93.93%, precision 89.40%, and F1-score 91.74%.
• The proposed method demonstrated strong classification performance.

Conclusions:

• The ensemble deep learning algorithm shows significant potential for NCP diagnosis.
• The method effectively improves the performance of deep neural networks in multi-category prediction tasks.
• This approach can assist medical personnel in achieving efficient and fast diagnoses.