Comparison of AI Semantic features and radiomics features for COVID-19 differentiation

The source code of BigBiGAN training for semantic feature extreaction, and radiomics feature extraction, and linear classifier and Lasso classifier used in the study.

Please note that all the source code is executed in the open Google Colab computing platform (BigBiGAN) and the open source Python environment (Pyradiomics) to ensure the reproducibility of this study.

To reproducibility, please execute the following code in the order mentioned in the paper.

• Detailed instruction (by the order) of using the code is as follows.

• 1. BigBiGAN_Learning: Training of the BigBiGAN framework for 120-dimensional AI sementic features extraction.
• 1. Pyradiomics for 120-dimensional radiomics features extraction.
• 1. Classifiers: The execution code of linear classifier and Lasso classifier. The output are the ROC curves and the prediction accuracy obtained by each classifier on your datasets.