Skip to main content

Table 5 Segmentation performance of different methods in terms of the Dice score and F1 score, evaluated on cremi_triplet

From: Sparse self-attention aggregation networks for neural sequence slice interpolation

Methods cremi_triplet A cremi_triplet B cremi_triplet C
  Dice F1 Dice F1 Dice F1
SepConv-L s [9] 0.547 0.546 0.425 0.426 0.458 0.458
DAIN-L s [12] 0.560 0.563 0.340 0.342 0.384 0.388
SSAN(Ours) 0.602 0.602 0.435 0.435 0.465 0.467
  1. We report the mean metrics of the membrane boundary on the ground truth and intermediate image synthesized by different methods. The proposed SSAN algorithm significantly surpasses other methods in these evaluation metrics
\