Skip to main content
. 2018 Nov 26;9:989. doi: 10.3389/fneur.2018.00989

Figure 2.

Figure 2

Overview of proposed method, showing the U-Net for segmenting core (Ŝc) and penumbra (Ŝcp), which are forwarded to the encoder E of the CAE for transforming them into low-dimensional shape space representations such that they can be linearly interpolated and decoded to output a follow-up estimate Ŝl.