Skip to main content
. 2021 Aug 7;125(8):1111–1121. doi: 10.1038/s41416-021-01511-w

Fig. 4. Comparisons of diagnostic performance by radiologists with and without AI assistance.

Fig. 4

The performance of AI model HCCNet and three experienced radiologists on the internal test set (a, b) and external test set (c, d). Three radiologists’ original performances are denoted by filled dots, and the performances with HCCNet assistance are denoted by hollow dots. Dashed lines connected paired performance points of each radiologist. The asterisk denoted the performance of our model in the ‘balanced performance’ setting.