Publication

Abstract : Pneumonia is one of the deadliest diseases, causing difficulties breathing and reducing oxygen consumption in lungs. It primarily affects children and seniors over the age of 65. It was observed that pneumonia caused a significant fraction of child fatalities. Early detection helps to cure with affordable medications. The commonly used diagnostic test is chest X-ray imaging since these x-rays are considerably inexpensive and quicker than other imaging modalities. The emergence of artificial intelligence simplified many activities, particularly the processing and categorization of images using deep learning convolutional networks. Therefore, in this paper, a model is trained utilizing a transfer learning approach to provide a rapid pneumonia detection system. The pre-trained network DenseNet201 with global average pooling layer was employed to validate some techniques, such as segmentation, enhancement, and augmentation. These experiments were conducted on the openly accessible RSNA pneumonia detection challenge dataset. The DenseNet201 with enhanced images achieved the highest results of 95.95% for accuracy, 95.13% precision, 86.41 for recall, and 90.56% for F1 score. This score ensures that this technique outperforms some of the existing techniques.