Deep Learning Classifiers for Multi-Class Classification of Medical Images using Chest X-Ray Scans

Coronavirus is not the only virus that can cause pneumonia, but pneumonia caused by COVID-19 is more likely tobe severe than other types of pneumonia. Pneumonia is a dangerous consequence that occurs when the virus enters the lung tissue of the lower respiratory tract. This can happen when the infection is inhaled. The image of the internal thoracic organs was obtained during an X-ray examination. The major limitation of the existing system is the degradation problem. In the existing system, the parameters considered for their experimentation are learning rate 0.001, batch size 7.0, 50 epochs and 0.5 dropouts with SGD optimizer. To overcome the degradation problem, we have considered the modified parameters as 0.0001 learning rate, batch size of 32, number of epochs as 100, dropouts as 0.6 and optimizer as Adam. The main goal of this examination is to classify four different classes: Covid-19, Lung opacity, viral pneumonia, and normal by using deep learning classifiers like the EfficientNetV2 family. The dataset used in this study consists of 4000 chest X-ray images. Classified, different deep learning classifiers are investigated using pre-trained CNN models such as efficient family. Accuracy, precision, recognition, F1 score, and AUC are all important metrics used to verify the efficiency of the model. The EfficientNetV2S achieved the highest classification accuracy for both classes COVID-19 class is 99.57% and for a normal class is 99.85%, The EfficientNetV2B1 achieved an accuracy of Lung Opacity is 99.57% and EfficientNetV2B1 achieved the highest accuracy of viral pneumonia is 99.71%. The proposed technique is a useful application in clinical practice and helps physicians to identify diseases from chest X-ray images. This enables physicians to help patients promptly.