Table 4.
Comparison of different segmentation methods that participated in the SCGM Segmentation Challenge6 against each of the four manual segmentation masks of the test set, reported here in the format: mean (std).
| JCSCS | DEEPSEG | MGAC | GSBME | SCT | VBEM | Proposed Method | |
|---|---|---|---|---|---|---|---|
| DSC | 0.79 (0.04) | 0.80 (0.06) | 0.75 (0.07) | 0.76 (0.06) | 0.69 (0.07) | 0.61 (0.13) | 0.85 (0.04) |
| MSD | 0.39 (0.44) | 0.46 (0.48) | 0.70 (0.79) | 0.62 (0.64) | 0.69 (0.76) | 1.04 (1.14) | 0.36 (0.34) |
| HSD | 2.65 (3.40) | 4.07 (3.27) | 3.56 (1.34) | 4.92 (3.30) | 3.26 (1.35) | 5.34 (15.35) | 2.61 (2.15) |
| SHD | 1.00 (0.35) | 1.26 (0.65) | 1.07 (0.37) | 1.86 (0.85) | 1.12 (0.41) | 2.77 (8.10) | 0.85 (0.32) |
| SMD | 0.37 (0.18) | 0.45 (0.20) | 0.39 (0.17) | 0.61 (0.35) | 0.39 (0.16) | 0.54 (0.25) | 0.36 (0.17) |
| TPR | 77.98 (4.88) | 78.89 (10.33) | 87.51 (6.65) | 75.69 (8.08) | 70.29 (6.76) | 65.66 (14.39) | 94.97 (3.50) |
| TNR | 99.98 (0.03) | 99.97 (0.04) | 99.94 (0.08) | 99.97 (0.05) | 99.95 (0.06) | 99.93 (0.09) | 99.95 (0.06) |
| PPV | 81.06 (5.97) | 82.78 (5.19) | 65.60 (9.01) | 76.26 (7.41) | 67.87 (8.62) | 59.07 (13.69) | 77.29 (6.46) |
| JI | 0.66 (0.05) | 0.68 (0.08) | 0.60 (0.08) | 0.61 (0.08) | 0.53 (0.08) | 0.45 (0.13) | 0.74 (0.06) |
| CC | 47.17 (11.87) | 49.52 (20.29) | 29.36 (29.53) | 33.69 (24.23) | 6.46 (30.59) | −44.25 (90.61) | 64.24 (10.83) |
For of fair comparison, the metrics are the same as used in the study6 and the results from other methods are replicated here, where we have: Dice similarity coefficient (DSC), mean surface distance (MSD), Hausdorff surface distance (HSD), skeletonized Hausdorff distance (SHD), skeletonized median distance (SMD), true positive rate (TPR), true negative rate (TNR), positive predictive value (PPV), Jaccard index (JI) and conformity coefficient (CC). In bold font, we represent the best-obtained results on each metric. We also note that MSD, HSD, SHD and SMD metrics are in millimeters and that lower values mean better results.