Multi-center sparse learning and decision fusion for automatic COVID-19 diagnosis

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Computational Biology

Background:

• The COVID-19 pandemic, caused by SARS-CoV-2, has resulted in a significant rise in severe pneumonia cases requiring hospitalization.
• Early and automated diagnosis of COVID-19 is critical for epidemic control and reducing patient mortality.

Purpose of the Study:

• To propose a joint multi-center sparse learning (MCSL) and decision fusion scheme for automatic COVID-19 diagnosis using chest CT images.
• To address data inconsistencies across multiple centers and improve the generalization performance of diagnostic models.

Main Methods:

• Chest CT images were converted to histogram of oriented gradient (HOG) images to mitigate multi-center data differences.
• A 3D convolutional neural network (3D-CNN) was used to extract features from 3D HOG image slices.
• A multi-center sparse learning (MCSL) method was employed to select discriminative features and train multi-center classifiers, followed by decision fusion.

Main Results:

• The proposed method demonstrated effectiveness in improving COVID-19 diagnosis performance on chest CT images from five centers.
• Experimental results showed that the MCSL and decision fusion scheme outperformed existing state-of-the-art methods.
• The approach successfully reduced structural differences in multi-center data, enhancing generalization.

Conclusions:

• The developed joint MCSL and decision fusion scheme offers a robust solution for automatic COVID-19 diagnosis from chest CT scans.
• This method shows significant potential for improving diagnostic accuracy and aiding in the management of the COVID-19 pandemic.
• The study highlights the efficacy of integrating sparse learning and deep learning for multi-center medical image analysis.