Table 1 Comparison with mean (standard deviation) of Dice score, HD95 score, IoU, Precision and Recall of the ischemic stroke lesion segmentation using 100% CT perfusion training dataset.

From: Noise-induced self-supervised hybrid UNet transformer for ischemic stroke segmentation with limited data annotations

Methods

Dice

HD95 (mm)

IOU

Precision

Recall

ERFNet60

  0.537 (0.225)

  16.180 (6.604)

  0.415 (0.205)

  0.776 (0.242)

  0.470 (0.218)

UShape61

  0.476 (0.253)

  22.066 (15.751)

   0.366 (0.216)

  0.574 (0.270)

  0.459 (0.291)

nnUnet59

  0.577 (0.222)

  19.689 (13.439)

  0.455 (0.210)

  0.729 (0.232)

  0.532 (0.237)

USSLNet62

  0.582 (0.205)

  16.987 (15.675)

  0.451 (0.192)

  0.689 (0.107)

  0.597 (0.271)

TransBTS63

  0.439 (0.207)

  23.564 (14.298)

  0.308 (0.202)

  0.677 (0.162)

  0.367 (0.239)

UNETR37

  0.469 (0.229)

  20.648 (13.146)

  0.337 (0.208)

  0.503 (0.258)

  0.496 (0.252)

HUT40

  0.609 (0.181)

  16.591 (13.373)

  0.479 (0.179)

  0.701 (0.161)

  0.577 (0.189)

HUT-NSS

 0.624 (0.169)

 12.197 (8.780)

 0.495 (0.169)

 0.753 (0.152)

  0.587 (0.178)

Back to article page