Table 2.
Medians (top), and means with standard deviations (bottom) for validation measures for all the methods and all the datasets. We highlight the best results in based on the median values. Among all datasets, our PCA model has the best median on Dice overlap scores and generally on surface distances. Exception is BEaST which achieves a lower maximum surface distances on the LPBA40 and the TBI datasets. In addition, our model also has the best mean and variance for the Dice overlap scores and the surface distances on most of these datasets.
Dataset: IBSR | |||||||
---|---|---|---|---|---|---|---|
PCA | ROBEX | BEaST* | MASS | BET | BSE | CNN | |
Dice(%) | 97.07 96.99±0.53 |
95.09 94.98±1.17 |
96.94 95.07±7.50 |
92.23 86.76±11.06 |
95.66 95.16±0.96 |
95.68 89.54±21.76 |
84.50 81.10±12.07 |
Avg Dist(mm) | 0.71 0.79±0.27 |
1.52 1.51±0.56 |
0.74 1.48±2.76 |
2.50 4.84±4.54 |
1.31 1.49±0.47 |
1.18 4.16±10.53 |
4.62 5.59±3.10 |
95% Dist(mm) | 2.83 2.84±0.43 |
4.18 4.50±1.58 |
3.00 5.19±9.79 |
8.73 13.66±11.81 |
3.61 4.22±1.39 |
5.00 12.27±20.83 |
20.05 22.25±9.41 |
Max Dist(mm) | 9.30 11.97±8.14 |
15.55 17.30±8.40 |
11.74 15.60±13.41 |
24.20 25.76±12.94 |
15.68 17.91±7.85 |
16.57 24.93±23.32 |
39.10 41.10±8.14 |
Sensitivity(%) | 99.06 98.99±0.46 |
99.47 99.33±0.54 |
98.98 96.70±10.12 |
98.77 98.52±0.77 |
99.57 99.09±0.93 |
97.08 88.68±22.86 |
74.87 70.96±15.89 |
Specificity(%) | 99.48 99.44±0.21 |
98.89 98.90±0.51 |
99.48 99.32±0.37 |
97.88 96.11±3.81 |
99.12 98.98±0.46 |
99.58 99.15±1.67 |
99.85 99.80±0.19 |
Dataset: LPBA40 | |||||||
PCA | ROBEX | BEaST | MASS | BET | BSE | CNN | |
Dice(%) | 97.41 97.32±0.42 |
96.74 96.74±0.24 |
96.80 94.25±15.29 |
97.08 97.03±0.57 |
96.92 96.70±0.78 |
96.77 96.29±2.26 |
95.80 95.70±0.74 |
Avg Dist(mm) | 0.76 0.79±0.12 |
0.97 0.97±0.07 |
0.97 3.36±14.83 |
0.85 0.88±0.21 |
0.97 1.06±0.27 |
0.92 1.11±0.81 |
1.34 1.39±0.37 |
95% Dist(mm) | 2.28 2.27±0.32 |
2.98 2.97±0.26 |
2.86 6.45±23.19 |
2.36 2.50±0.97 |
3.46 3.92±1.24 |
2.62 3.46±3.38 |
3.11 3.56±1.56 |
Max Dist(mm) | 9.73 10.83±3.76 |
12.89 13.81±3.47 |
8.80 12.89±24.44 |
10.71 11.53±4.04 |
14.44 15.14±3.75 |
13.61 15.54±7.74 |
16.91 19.55±8.17 |
Sensitivity(%) | 97.10 96.81±1.23 |
96.15 96.33±0.85 |
94.76 92.70±15.12 |
95.05 95.15±1.08 |
98.70 98.66±0.54 |
94.85 94.02±4.10 |
92.79 92.62±1.46 |
Specificity(%) | 99.62 99.61±0.16 |
99.54 99.49±0.16 |
99.80 99.70±0.21 |
99.86 99.83±0.12 |
99.09 99.04±0.34 |
99.81 99.79±0.09 |
99.85 99.83±0.07 |
Dataset: BRATS | |||||||
PCA | ROBEX | BEaST* | MASS | BET | BSE | CNN | |
Dice(%) | 96.34 96.16±0.92 |
95.15 94.83±1.49 |
94.99 93.29±7.00 |
96.20 95.71±1.39 |
94.40 90.95±13.41 |
84.21 84.91±8.89 |
1.75 21.89±29.54 |
Avg Dist(mm) | 0.97 1.00±0.31 |
1.31 1.54±0.70 |
1.38 2.28±3.34 |
1.03 1.17±0.56 |
1.68 7.58±25.30 |
4.65 4.37±3.61 |
55.88 44.87±29.05 |
95% Dist(mm) | 4.15 4.35±1.27 |
5.23 6.03±2.50 |
5.01 7.71±9.29 |
4.62 4.87±2.04 |
5.87 6.18±3.53 |
13.52 13.92±13.00 |
78.45 73.85±38.77 |
Max Dist(mm) | 12.03 15.26±9.32 |
12.83 16.42±8.80 |
15.03 20.32±14.57 |
13.29 16.43±8.97 |
17.49 22.78±22.61 |
33.95 32.02±22.38 |
87.73 86.60±36.92 |
Sensitivity(%) | 96.16 96.17±1.84 |
95.82 94.95±2.88 |
97.45 97.06±2.66 |
94.46 93.62±2.87 |
94.92 94.77±3.82 |
74.28 77.80±13.43 |
0.89 16.17±24.73 |
Specificity(%) | 99.33 99.29±0.25 |
99.16 98.98±0.65 |
98.68 97.28±5.46 |
99.73 99.62±0.28 |
98.90 93.69±22.08 |
99.82 99.29±2.38 |
100.00 99.97±0.05 |
Dataset: TBI | |||||||
PCA | ROBEX | BEaST | MASS | BET | BSE | CNN | |
Dice(%) | 96.40 96.28±0.85 |
93.71 93.60±1.00 |
95.04 95.06±0.96 |
95.11 95.42±0.96 |
95.40 95.14±1.12 |
92.64 91.00±4.31 |
91.51 90.40±5.07 |
Avg Dist(mm) | 1.20 1.22±0.30 |
2.12 2.20±0.40 |
1.60 1.62±0.33 |
1.61 1.53±0.33 |
1.50 1.57±0.40 |
2.23 3.15±1.66 |
3.20 3.46±1.75 |
95% Dist(mm) | 3.37 3.41±0.85 |
6.20 5.99±0.97 |
6.24 5.86±1.44 |
4.90 4.59±0.77 |
5.06 5.57±1.91 |
9.46 13.07±7.11 |
14.97 16.04±8.72 |
Max Dist(mm) | 16.13 15.54±5.03 |
18.25 18.89±5.12 |
15.09 17.27±4.60 |
15.65 16.08±4.16 |
17.46 18.53±4.59 |
27.16 31.96±12.71 |
41.49 37.06±10.09 |
Sensitivity(%) | 97.82 97.76±0.92 |
98.91 98.64±0.93 |
92.12 92.89±2.44 |
98.74 98.68±0.66 |
93.98 93.65±1.87 |
87.51 85.65±8.17 |
85.69 83.77±8.58 |
Specificity(%) | 99.03 99.07±0.26 |
98.09 97.75±0.93 |
99.63 99.57±0.22 |
98.64 98.59±0.47 |
99.41 99.44± 0.11 |
99.82 99.54±0.86 |
99.80 99.81±0.07 |