Table 7.
The comparison of the accuracy of MAPS, BET, BSE and HWA. The table shows the differences in the median (95% CI) of Jaccard index, false positive rate and false negative rate between the four automated brain extraction methods.
| Jaccard index (using thresholded segmenta- tions) |
False positive rate / % (using thresholded seg- mentations) |
False negative rate / % | ||
|---|---|---|---|---|
| 1.5T | ||||
| MAPS vs BET | 0.012* (0.011, 0.013) | −0.016* (−0.022, −0.009) | −0.792* (−0.876, −0.724) | |
| MAPS vs BSE | 0.028* (0.021, 0.038) | 0.068* (0.058, 0.078) | −2.16* (−3.09, −1.57) | |
| MAPS vs HWA | 0.011* (0.009, 0.012) | −0.113* (−0.122, −0.102) | 0.002 (−0.001, 0.004) | |
| HWA vs BET | 0.001 (−0.000, 0.003) | 0.097* (0.086, 0.105) | −0.793* (−0.878, −0.726) | |
| HWA vs BSE | 0.018* (0.010, 0.028) | 0.181* (0.169, 0.192) | −2.16* (−3.09, −1.57) | |
| BET vs BSE | 0.016* (0.009, 0.026) | 0.084* (0.075, 0.095) | −1.37* (−2.34, −0.807) | |
| 3T | ||||
| MAPS vs BET | 0.015* (0.012, 0.018) | 0.015* (0.000, 0.030) | −1.28* (−1.52, −1.17) | |
| MAPS vs BSE | 0.079* (0.072, 0.086) | 0.102* (0.086, 0.117) | −9.18* (−10.0, −8.64) | |
| MAPS vs HWA | 0.018* (0.015, 0.021) | −0.144* (−0.184, −0.114) | 0.008* (0.003, 0.015) | |
| HWA vs BET | −0.003 (−0.007, 0.001) | 0.159* (0.131, 0.199) | −1.29* (−1.53, −1.18) | |
| HWA vs BSE | 0.062* (0.055, 0.068) | 0.246* (0.220, 0.285) | −9.19* (−10.0, −8.65) | |
| BET vs BSE | 0.065* (0.058, 0.072) | 0.087* (0.072, 0.106) | −7.90* (−8.77, −7.29) | |
*
denotes statistical significance at p < 0.05.