Covid-19 Detection on Chest X-ray Images Using Semi-supervised Deep Learning with MixMatch and CLAHE

The COVID-19 pandemic has caused a global crisis, emerging as one of the most critical diseases in history. The prevalence of COVID-19 continues to increase worldwide. The real-time polymerase chain reaction (RT-PCR) test is internationally recognized for diagnosing COVID-19 in suspected cases. Despite being the diagnostic gold standard, RT-PCR is time-consuming and expensive, with a significant false negative rate. Deep learning techniques have shown promise in the automated diagnosis of COVID-19. However, these methods typically require large labeled datasets, which are difficult to obtain during virus outbreaks. This paper introduces a semi-supervised deep learning model that addresses this issue with labeled and unlabeled data. We developed, assessed, and refined a semi-supervised deep learning framework based on the MixMatch architecture to classify chest X-rays as COVID-19 or typical cases. In addition, a contrast-limited adaptive histogram equalization (CLAHE) method is applied to improve the visibility of small details, texture, and local contrast in chest X-ray images. Our proposed approach achieves classification performance with 99.8% accuracy and 99.8% F1 Score.