Table 5.
w/o fine-tuning subnet | w/o weighted voting subnet | w/o atlas mask | DLF | ||
---|---|---|---|---|---|
Dice similarity coefficient | Brain-7T-T2 dataset | ||||
GDSC | 61.3±6.9 | 61.4±10.5 | 70.7±3.8 | 71.2±4.3 | |
Hippo | 81.0±9.2 | 81.7±13.2 | 90.8±1.4 | 90.7±1.5 | |
CA1 | 58.8±14.9 | 65.0±13.4 | 75.5±3.4 | 75.5±3.9 | |
CA2 | 38.1±24.5 | 57.0±14.0 | 67.1±6.4 | 67.8±6.4 | |
CA3 | 49.1±16.5 | 56.4±15.5 | 69.3±5.5 | 69.6±5.6 | |
DG | 67.2±18.4 | 70.3±15.0 | 80.9±2.0 | 80.6±2.6 | |
SUB | 71.9±6.7 | 67.8±9.7 | 72.0±5.4 | 73.4±5.7 | |
TAIL | 73.7±5.6 | 68.5±13.9 | 80.5±2.8 | 80.4±3.3 | |
ERC | 68.8±8.7 | 61.6±18.8 | 80.2±5.0 | 79.5±5.7 | |
BA35 | 54.4±10.1 | 56.3±14.6 | 65.2±9.5 | 66.0±9.8 | |
BA36 | 57.0±10.3 | 56.3±15.5 | 66.2±7.9 | 67.5±8.7 | |
PHC | 66.6±6.5 | 69.1±11.3 | 72.8±9.3 | 72.4±8.2 | |
Lumbar-Healthy dataset | |||||
Lumbar | 79.9±5.6 | 78.1±5.9 | 83.0±4.7 | 83.4±4.7 | |
Mean surface distance (mm) | Brain-7T-T2 dataset | ||||
Hippo | 0.65±0.26 | 0.70±0.58 | 0.38±0.06 | 0.38±0.06 | |
CA1 | 0.91±0.38 | 1.16±1.36 | 0.63±0.11 | 0.61±0.09 | |
CA2 | 1.73±1.91 | 0.88±0.87 | 0.56±0.14 | 0.54±0.13 | |
CA3 | 1.17±0.93 | 1.05±1.13 | 0.55±0.11 | 0.54±0.11 | |
DG | 0.77±0.42 | 0.97±1.09 | 0.55±0.07 | 0.57±0.09 | |
SUB | 0.78±0.23 | 0.93±0.30 | 0.84±0.21 | 0.78±0.16 | |
TAIL | 0.69±0.12 | 0.94±0.67 | 0.57±0.09 | 0.57±0.09 | |
ERC | 0.82±0.20 | 1.13±0.88 | 0.61±0.18 | 0.60±0.19 | |
BA35 | 1.10±0.28 | 1.15±0.68 | 0.95±0.27 | 0.94±0.28 | |
BA36 | 1.31±0.33 | 1.38±0.69 | 1.09±0.21 | 1.05±0.22 | |
PHC | 0.92±0.28 | 0.78±0.26 | 0.83±0.31 | 0.87±0.32 | |
Lumbar-Healthy dataset | |||||
Lumbar | 2.63±0.46 | 2.44±0.37 | 2.16±0.50 | 2.12±0.50 |
Abbreviations: DLF = deep label fusion; Hippo = hippocampus; AHippo/PHippo = anterior/posterior hippocampus; CA1–3 = cornu ammonis 1 to 3; dentate gyrus = DG; subiculum = SUB; TAIL = the tail of hippocampus; ERC = entorhinal cortex; BA35/36 = Brodmann areas 35/36; PHC = parahippocampal cortex; w/ and w/o = with and without.