Skip to main content
. 2022 Jun;12(6):3276–3287. doi: 10.21037/qims-21-1089

Figure 6.

Figure 6

Performance of the multivariate classification deep learning models (PM3: VGGNet-16-based pretraining model; PM4: ResNet-50-based pretraining model; AM3: VGGNet-16-based AT model; AM4: ResNet-50-based AT model) in the internal validation datasets (left) and test sets (right). The kappa calculation results range from −1 to 1, but κ ranges usually between 0 and 1 and is further divided into five groups to express different consistency levels (0–0.20: slight consistency, 0.21–0.40: fair consistency, 0.41–0.60: moderate consistency, 0.61–0.80: substantial consistency, and 0.81–0.99: almost perfect consistency). Error bars show the standard error of κ values. AT, adversarial training; ResNet, residual network; VGGNet, Visual Geometry Group network.