Study of SQ-Net behavior using t-SNE activation by Classification of Lung Diseases in Chest Radiographs.

Due to population ageing, urbanisation, and the risks connected to these factors, lung diseases will increasingly be a leading cause of death and morbidity. Lung conditions are increasingly being acknowledged as a major cause of death on a global scale. Therefore, it is crucial to address risk factors, increase vaccination rates, and enhance access to oxygen in order to lower lung disease mortality globally. This is important because, according to medical professionals, respiratory experts, and epidemiologists, a lack of representative estimates of the burden of lung disease is a problem that could limit the adoption of cost-effective, evidence-based strategies in the future. This inspires us to classify lung diseases such as Covid19, pneumonia, normal, tuberculosis, and pneumothorax using the deep learning model SQ-Net on chest radiographs. In order to study and comprehend the network behaviour in lung disease classification, we also applied G-CAM and t-SNE (t-Stochastic Neighbor Embedding) activation mapping to the max-pooling, final convolutional, and softmax layers.