Attention-COVIDNet: An Enhanced COVID-19 Classification Framework Using Attention Mechanism

The COVID-19 pandemic has socio-economically affected world health and economies, thus increasing the urgency for diagnostic procedures that are characterized by efficiency, accuracy, and cost-effectiveness. In attendance to this difficulty, we present Attention-COVIDNet as a very special deep learning framework for the automatic classification of COVID-19 from chest X-ray pictures. Some of the advanced image preprocessing techniques used by the proposed model are histogram equalization and region of interest extraction. Basically, these techniques are used to enhance image quality and focus on more important features. The classification will be done with the help of an attention-based Convolutional Neural Network, which will be changing the importance given to different portions of the X-ray images dynamically. This improves distinguishment in the model for COVID-19, pneumonia, and healthy instances. Attention-COVIDNet was trained on a comprehensive dataset of chest X-ray images and reached 98.2% accuracy with an Fl-score of 98.5%, hence outperforming state-of-the-art models in accuracy. By introducing the attention mechanism to the CNN design, diagnostic precision is enhanced. This enables the model to extract subtle patterns that indicate COVID-19. Attention-COVIDNet has the potential to turn into a dependable tool to assist health care providers for the early and accurate diagnosis of COVID-19, which eventually will lead to better outcomes for patients and also in the form of better treatment of the pandemic. The findings from this study enhance the promise of Attention-COVIDNet.