Abstract:
Despite the fact that pneumonia is a vaccine-preventable disease, it has become a global
health concern. Every year, 800,000 children die from pneumonia. Pneumonia has been
the cause of death in 17 percent of Bangladeshi kids under the age of 5. X-rays of the chest
can be used to detect it. A qualified radiologists is essential for analysis. Even with an
experienced radiologist, analyzing chest X-rays is tough. Machines can readily complete
jobs that are difficult for humans to complete. Pneumonia can be diagnosed using machine
learning and image processing. Machine learning (ML) is a branch of artificial intelligence
that helps computers learn and make decisions on their own. Efficient pretrained transfer
learning based VGG16, VGG19 and ResNet50 architectures are used to develop the
proposed model which could assist both doctors and patients. The convolutional neural
network is one of the machine learning algorithms. Image processing has also been used to
help computers comprehend photos and identify Pneumonia more precisely. We used
augmentation approaches even after training our model with over 6,000 pictures. With the
use of a chest x-ray report, users will be able to quickly determine if they have Pneumonia
in a relatively short period of time. In order to improve the accuracy, we added several
extra layers to all the model. To enhance the number of images in a balanced way, data
augmentation techniques were applied. The comparison between VGG16, VGG19, and
ResNet50 is shown in this project. Whereas, VGG16 has obtained 94.5 percent accuracy,
which is the greatest among the other algorithms implemented, indicating that it can
classify normal and pneumonia. Other image-based classification problems can also apply
the proposed architecture.
