Table 1.

Pseudocode of the optimal backbone model selection algorithm.

Step1:	load the COVID-19 dataset.
Step2:	load the pre-trained backbone models, including AlexNet, ResNet-18, ResNet-50, MobileNetV2, and EfficientNet.
Step3:	modify the structure of these backbones based on the labels of the COVID-19 dataset using ELM.
Step4:	train these models and test them based on 5-fold cross-validation.
Step5:	compute the average testing accuracies of the 5 backbones.
Step6:	output the optimal backbone model which yielded the highest average testing accuracy.