Covid-19 classification by FGCNet with deep feature fusion from graph convolutional network and convolutional neural...

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Infectious Diseases

Background:

• COVID-19 is a global infectious disease requiring rapid diagnostic tools.
• Chest CT imaging is a key modality for COVID-19 diagnosis.
• Current diagnostic methods can be time-consuming.

Purpose of the Study:

• To develop an artificial intelligence (AI) based tool for COVID-19 diagnosis using chest CT images.
• To enhance diagnostic accuracy and speed for radiologists.
• To create a novel deep learning model for improved COVID-19 detection.

Main Methods:

• Developed a novel deep learning model named FGCNet.
• Utilized a convolutional neural network (CNN) for image-level feature extraction, incorporating rank-based average pooling and multiple-way data augmentation.
• Employed a graph convolutional network (GCN) to learn relation-aware representations.
• Implemented deep feature fusion (DFF) to integrate features from CNN and GCN.

Main Results:

• FGCNet was selected as the best performing model among eight proposed network architectures.
• The FGCNet model demonstrated superior performance compared to 15 state-of-the-art methods.
• Achieved high accuracy in detecting COVID-19 from chest CT images.

Conclusions:

• The proposed FGCNet model is effective for COVID-19 detection.
• FGCNet offers improved performance over existing state-of-the-art approaches.
• FGCNet can serve as a valuable tool to assist radiologists in rapid COVID-19 detection from chest CT scans.