Skip to main content
. 2020 Jan 15;19:4. doi: 10.1186/s12938-020-0748-9

Fig. 5.

Fig. 5

Representative slices of the averaged absolute error before and after harmonization by using different methods. Note: dual generative adversarial networks are performed with the 2-dimensional kernel on axial slices. Absolute errors after harmonization have been averaged over different runs of the sixfold cross-validation