Table 5 Skull stripping performance comparison to state-of-the-arts.
From: A general skull stripping of multiparametric brain MRIs using 3D convolutional neural network
| Â | # of cases | Dice | Precision | Recall | FPR | FNR | HD95 (mm) | p-value (t-test on Dice) |
|---|---|---|---|---|---|---|---|---|
BET23 | 216 | 0.8494 ± 0.0455 | 0.7463 ± 0.0718 | 0.9916 ± 0.0186 | 0.0650 ± 0.0224 | 0.0084 ± 0.0186 | 19.9951 ± 5.4941 | \(9.19\times {10}^{-151}\) |
3dSS47 | 216 | 0.8427 ± 0.0449 | 0.7430 ± 0.0751 | 0.9809 ± 0.0279 | 0.0660 ± 0.0238 | 0.0191 ± 0.0279 | 19.9087 ± 4.4316 | \(1.29\times {10}^{-159}\) |
ROBEX48 | 216 | 0.9555 ± 0.0173 | 0.9730 ± 0.0236 | 0.9396 ± 0.0318 | 0.0053 ± 0.0057 | 0.0604 ± 0.0318 | 4.4792 ± 1.8869 | \(7.52\times {10}^{-54}\) |
UNet3D38 | 216 | 0.9773 ± 0.0179 | 0.9818 ± 0.0168 | 0.9735 ± 0.0290 | 0.0035 ± 0.0034 | 0.0265 ± 0.0290 | 3.1219 ± 2.7262 | \(5.61\times {10}^{-6}\) |
UPNN39 | 216 | 0.9743 ± 0.0257 | 0.9814 ± 0.0156 | 0.9684 ± 0.0405 | 0.0035 ± 0.0032 | 0.0316 ± 0.0405 | 3.3924 ± 2.8434 | \(4.22\times {10}^{-7}\) |
EnNet (ours) | 216 | 0.9850 ± 0.0171 | 0.9940 ± 0.0093 | 0.9768 ± 0.0307 | 0.0012 ± 0.0019 | 0.0232 ± 0.0307 | 2.6098 ± 2.4814 | – |
