Deep Learning Based Approaches to Detect Covid-19 with X-Ray Images

Publication Name

2022 IEEE Asia-Pacific Conference on Image Processing, Electronics and Computers, IPEC 2022


In 2019, the outbreak of COVID-19 impacted on people's life and the economy around the world. In medical science circle, scientists invented and utilized a reverse transcription polymerase chain reaction (RT-PCR) to search for this infectious disease. However, due to the hidden disease that spreads fast, the method of RT-PCR did not test for a large number of people in a short period. Therefore, it is urgent for researchers to find an alternation as a method for diagnosis. Deep learning is a crucial method to replace the original way in the medical imaging. The enough images are critical deficiency in the beginning of the outbreak of COVID-19. Moreover, we collect tomography (CT) images and X-ray images from different websites to detect COVID-19. Initially, this method was used to prepare a database of two classes: COVID-19 and Normal, and then comparing different DL models to search for two models that have better performance. Now, the usage of two models could be applied in three classes classification: Pneumonia, COVID-19 and Normal. The two deep learning models we choose (ResNet50 and VGG16) are fine-tuned and trained through transfer learning. The final result is 95.6% of overall accuracy. The accuracy is the most accurate by comparing the different classification algorithms.

First Page


Last Page




Link to publisher version (DOI)