Deep learning-based multi-view fusion model for screening 2019 novel coronavirus pneumonia: A multicentre study

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Radiology

Background:

• Coronavirus disease 2019 (COVID-19) poses a significant global health challenge.
• Accurate and rapid diagnosis of COVID-19 pneumonia is crucial for patient management and public health.
• Computed Tomography (CT) imaging is a valuable tool for diagnosing COVID-19 pneumonia.

Purpose of the Study:

• To develop and evaluate a deep learning-based method for assisting radiologists in the rapid and accurate identification of COVID-19 pneumonia using CT images.
• To leverage multi-view CT data for enhanced diagnostic performance.

Main Methods:

• A retrospective collection of 495 chest CT datasets from three Chinese hospitals.
• Random division of datasets into training (80%), validation (10%), and testing (10%) sets.
• Development of a multi-view fusion deep learning model trained on axial, coronal, and sagittal CT views to screen for COVID-19.

Main Results:

• The model achieved an Area Under the Curve (AUC) of 0.732, accuracy of 0.700, sensitivity of 0.730, and specificity of 0.615 in the validation set.
• In the testing set, the model demonstrated improved performance with an AUC of 0.819, accuracy of 0.760, sensitivity of 0.811, and specificity of 0.615.
• The multi-view deep learning approach showed robust performance in differentiating COVID-19 pneumonia.

Conclusions:

• The proposed deep learning model trained on multi-view chest CT images shows significant potential for improving diagnostic efficacy.
• The AI-driven method can assist radiologists in the initial screening of COVID-19 pneumonia, potentially mitigating heavy workloads.
• This approach highlights the value of deep learning in medical image analysis for infectious disease diagnosis.