Table 2.
The performance of hippocampal segmentation on three datasets. Paired t‐tests were performed for comparing the LML method with others. Dice similarity coefficients (median ± standard deviation) along with statistical results for both left and right hippocampus are reported
| Dataset | Method | Left hippocampus | Right hippocampus | ||||
|---|---|---|---|---|---|---|---|
| Dice | t‐stat | P‐value | Dice | t‐stat | P‐value | ||
| In‐house | LMV | 0.879 ± 0.057 | 5.2 | 1.8e‐5 | 0.860 ± 0.048 | 7.8 | 2.3e‐8 |
| LWINV | 0.896 ± 0.031 | 4.1 | 3.4e‐4 | 0.895 ± 0.039 | 3.9 | 5.2e‐4 | |
| LML | 0.902 ± 0.026 | – | – | 0.900 ± 0.034 | – | – | |
| ADNI 1.5T | LMV | 0.841 ± 0.083 | 9.1 | 9.1e‐13 | 0.833 ± 0.101 | 8.5 | 9.5e‐12 |
| LWINV | 0.886 ± 0.040 | 4.2 | 1.0e‐4 | 0.886 ± 0.044 | 4.8 | 1.0e‐5 | |
| LML | 0.891 ± 0.033 | – | – | 0.891 ± 0.036 | – | – | |
| ADNI 3.0T | LMV | 0.817 ± 0.087 | 7.7 | 1.7e‐9 | 0.819 ± 0.080 | 7.7 | 1.6e‐9 |
| LWINV | 0.876 ± 0.029 | 4.5 | 6.3e‐5 | 0.878 ± 0.025 | 4.2 | 1.2e‐4 | |
| LML | 0.881 ± 0.022 | – | – | 0.882 ± 0.019 | – | – | |
LWINV, local weighted inverse distance voting; LMV, local majority voting; LML, local manifold learning.