Table 3 Performance of the multi-arm U-Net on the test set of the ISBI 2015 challenge.

From: Multi-arm U-Net with dense input and skip connectivity for T2 lesion segmentation in clinical trials of multiple sclerosis

Team

Score

DC

PPV

LTPR

LFPR

AVD

Best model

93.36

0.669

0.886

0.538

0.125

0.396

Zhang et al.2

93.21

0.643

0.908

0.520

0.124

0.428

Ours (5.9 M)

93.04

0.677

0.865

0.520

0.143

0.362

Brugnara et al.5* (19.1 M)

93.03

0.681

0.855

0.538

0.157

0.366

Hashemi et al.28

92.48

0.584

0.921

0.414

0.087

0.497

Aslani et al.3

92.12

0.611

0.899

0.410

0.139

0.454

  1. The highest or lowest values are in bold.
  2. *3D nnU-Net trained for MS T2 lesion segmentation. The number of model parameters for multi-arm and nnU-Net is provided within parenthesis.
  3. AVD absolute volume difference, DC dice coefficient, LFPR lesion false positive rate, LTPR lesion true positive rate, PPV positive predictive values.
Back to article page