Transfer learning-based approach for detecting COVID-19 ailment in lung CT scan

Comput Biol Med. 2021 Aug;135:104575. doi: 10.1016/j.compbiomed.2021.104575. Epub 2021 Jun 12.


This research work aims to identify COVID-19 through deep learning models using lung CT-SCAN images. In order to enhance lung CT scan efficiency, a super-residual dense neural network was applied. The experimentation has been carried out using benchmark datasets like SARS-COV-2 CT-Scan and Covid-CT Scan. To mark COVID-19 as positive or negative for the improved CT scan, existing pre-trained models such as XceptionNet, MobileNet, InceptionV3, DenseNet, ResNet50, and VGG (Visual Geometry Group)16 have been used. Taking CT scans with super resolution using a residual dense neural network in the pre-processing step resulted in improving the accuracy, F1 score, precision, and recall of the proposed model. On the dataset Covid-CT Scan and SARS-COV-2 CT-Scan, the MobileNet model provided a precision of 94.12% and 100% respectively.

Keywords: COVID-19; CT Scan; Deep learning; MobileNet; Pandemic; Transfer learning.

MeSH terms

  • COVID-19* / diagnostic imaging
  • Deep Learning*
  • Humans
  • Lung* / diagnostic imaging
  • Tomography, X-Ray Computed