Fig. 2.
When comparing the performance of human readers and DLS on the validation data set, the complimentary performance of the DLS with high sensitivity (p < 0.05 for all readers) but lower specificity is already evident (AUC reader 1 [red] = 0.60, reader 2 [green] = 0.77, reader 3 [dark blue] = 0.75 and reader 4 [light blue] = 0.74). Readers 1 to 4 represent radiologists from highest to lowest levels of experience, respectively.