Skip to main content
. Author manuscript; available in PMC: 2021 Mar 1.
Published in final edited form as: IEEE Trans Med Imaging. 2020 Mar 5;39(9):2881–2892. doi: 10.1109/TMI.2020.2972547

Fig. 7.

Fig. 7

Histograms of per-voxel KLD associated to the uncertainty estimates as measured by the sample variance, shown for models trained with different augmentations on a) severe artefacts and b) minor artefacts. Note that the clean and classically augmented models produce a much higher number of estimates with high KLD and high variance (bottom-right corner of the histograms) compared to the motion augmented model.