부산대학교 도서관

The novel coronavirus disease has produced destructive effects on human life, taking away millions of lives. The biggest bottleneck in detecting the COVID-19-affected patient is the limited availability and time-consuming features of conventional RT-PCR tests and the lack of specialized sample extraction laboratories. Early detection of this virus may help in the advancement of a medication approach and disease control strategies. In this research, we have developed an Android smartphone application that can detect pneumonia and COVID-19 from chest X-ray photographs using convolutional neural network deep learning algorithms (VGG16 and VGG19). The COVID-19, pneumonia, and healthy chest X-ray images are collected from various repositories of a public database, Kaggle. After applying the data augmentation technique, 9,000 chest X-ray photographs were used for training, including 3,000 images for COVID-19, pneumonia, and normal cases. For testing, 3,000 chest X-ray photographs were collected, with 1,000 images for all three cases. VGG16 model achieved better performance than the VGG19 with a training accuracy of 98.31% and validation accuracy of 95.03%. Next, the deep learning-based automatic classification framework is deployed into a smartphone application. Finally, the application has been tested and assessed by a focused group, and analytical results have been presented.