Abstract: The COVID-19 pandemic, caused by the SARS-CoV-2 virus, has led to a global health crisis with significant morbidity and mortality. Effective screening methods are crucial for controlling its spread, but existing pathological tests have limited accuracy. Chest radiography imaging, including X-rays and CT scans, offers an adjunctive screening approach, yet interpretation challenges persist due to subtle markers and similarities with other pulmonary diseases. Deep learning architectures, particularly convolutional neural networks (CNNs), present a promising avenue for enhancing diagnostic accuracy. This study explores the modification of the VGG16 model with an attention layer to improve COVID-19 detection from chest X-ray images. The attention layer highlights relevant features, aiding in the identification of infection markers. Additionally, the model is extended to estimate infection severity, enhancing its diagnostic capabilities. Performance evaluation demonstrates promising results, suggesting the potential impact of attention-based modifications in refining existing architectures for improved COVID-19 screening. This research contributes to the evolving landscape of using deep learning models for COVID-19 detection and severity estimation, offering insights for future research and applications.
Keywords: Mean squared error (MSE), mean absolute error (MAE), Acute Respiratory Distress Syndrome (ARDS),convolutional neural networks (CNNs).
Copyright Form