CODENET: A deep learning model for COVID-19 detection

Hong Ju, Yanyan Cui, Qiaosen Su, Liran Juan, Balachandran Manavalan

Abstract: Highlights•CodeNet, a new deep learning method, addresses limitations of traditional testing by using chest X-rays (CXRs) for faster and cheaper diagnosis.•CodeNet achieves a remarkable 94.20% accuracy on the evaluation dataset, surpassing five existing methods.•CodeNet leverages contrastive learning to effectively extract features and generalize across different data sets, maximizing the use of latent information in CXR images.•Ablation studies and interpretability analysis demonstrates CodeNet's potential to guide medical professionals.