Figure 3
Comparison of the recovered water for the PEFC_1 dataset. The first column presents the first time step (T1), the second column time step 15 (T2) and the third column time step 30 (T3). (a) Difference volumes between the manual segmentation (ground truth) and SIRT-FBP-MS-D-DIFF water classifications with standard (Stand.), quasi-4D (Q.-4D) and ensemble (Ens.) training schemes of the MS-D network, respectively. The green color corresponds to features present only in the ground truth, orange to features present only in the SIRT-FBP-MS-D-DIFF water classifications. (b) Difference slice (slice 143, T3) between the water classification with the networks and ground truth segmentation. The white structures correspond to features present only in the ground truth segmentation, black structures only in the SIRT-FBP-MS-D-DIFF water classification.
