Fig. 3
From: Deep learning for detection of radiographic sacroiliitis: achieving expert-level performance
Sensitivity and 1-specificity (false positive rate) on the test and validation datasets using different cut-off values for the model predictions regarding the presence of definite radiographic sacroiliitis (classification as non-radiographic or radiographic axial spondyloarthritis). We analysed three cut-off values, indicated by vertical dashed lines. Cut-off 1 weights sensitivity over specificity, cut-off 2 weights specificity over sensitivity and cut-off 3 aims to be the optimal balance between the two performance measures. Cut-offs were only calculated on the validation dataset and then applied to the test and validation datasets