Skip to main content
. 2022 May 8;146:105587. doi: 10.1016/j.compbiomed.2022.105587

Fig. 2.

Fig. 2

The COVID-19 Pre-trained model architecture using machine-driven design exploration [3].novel coronavirus pneumonia, common pneumonia, and normal controls. The trained model achieves an overall accuracy of 99.1% with COVID-19 sensitivity (or recall) of 97.3% on the test data (20% of the data) which comprises of 120 COVID-19, 120 pneumonia, and 50 normal patient cases. The specificity of the model for COVID-19 is 99.9% and the positive predictive value (or precision) is 99.7%.