An Efficient Approach to detect the Pneumonia Using Advanced Deep Learning Techniques

Pneumonia is a lung infection that mostly affects people in developing and underdeveloped countries. Chest X-rays are the most common way to diagnose this condition. Even an experienced radiologist finds it challenging to examine chest X-rays. A novel deep learning model UNET with MobileNet, ResNet, and DenseNet121 as the backbone is designed to improve diagnosis accuracy, which will help radiologists to make a more accurate prediction of disease. An ensemble of three deep learning models is developed to make the work of doctors simpler. The quality of the model’s results is proportional to the quantity of data utilised for training. Accuracy, area under the curve (AUC), F1 score, and loss are only few of the metrics used to assess the model’s efficacy in this context. When compared to state-of-the-art approaches, the proposed method reliably predicts pneumonia from an X-ray image that has not been seen. These models are designed to predict the object class and the exact area of the lung opacity in the X-ray image and compared with ground truth mask.