Mean-Teacher-Model-with-Consistency-Regularization-for-Semi-Supervised-Detection-of-COVID-19

This study provides a novel approach for detecting COVID-19 using cough recordings. The method employs a Mean Teacher model with consistency regularization, utilizing both labeled and unlabeled data. The model is made up of a student network and a teacher network, with the teacher network guiding the training of the student network. Generalization is improved by maintaining consistency in the predictions of the student network. We used a collection of cough recordings from COVID-19 patients and healthy people, with balanced labeled and unlabeled parts. Accuracy, loss, precision, recall, and AUC are all measured using k-fold cross-validation. The results suggest that semi-supervised COVID-19 identification can be successful. When labeled data is scarce, the findings highlight the value of cough recordings and semi-supervised learning.