Comparison of transfer learning method for COVID-19 detection using convolution neural network

Helmi Imaduddin, Fiddin Yusfida Ala, Azizah Fatmawati, Brian Aditya Hermansyah

Abstract


Currently, one of the most dangerous diseases is Coronavirus disease 2019 (COVID-19). COVID-19 is a threat to the whole world, and almost all countries are experiencing the COVID-19 pandemic, including Indonesia. Various ways to detect COVID-19 sufferers have been carried out, such as swab tests, rapid tests, and antigens. One way that can be done to detect COVID-19 infection is to look at X-ray images of the patient's lungs because someone infected with COVID-19 has a different lung shape from normal people. Many studies have been carried out to detect COVID-19, using either machine learning (ML) or deep learning (DL). In this study, we propose to use transfer learning as an extraction feature in the classification of the covid dataset. The study was conducted four times using four different methods, namely ResNet 50, MobileNet V2, Inception V3, and DensNet-201. After experimenting, we compared the results to find out which method has the best results in detecting COVID-19. From this research, it was found that the ResNet 50 model has the best results with 92.3% accuracy, 93% precision, 93% F1-Score, 99% sensitivity, and 90.7% specificity.

Keywords


CNN; Covid-19; Deep Learning; Transfer Learning

Full Text:

PDF


DOI: https://doi.org/10.11591/eei.v11i2.3525

Refbacks

  • There are currently no refbacks.


Creative Commons License
This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.

Bulletin of EEI Stats