Table 2. Accuracy measures for segmenting three muscle compartments using the proposed method, ANTs and STAPLE.
| SCI Group | ND Group | |||||||
|---|---|---|---|---|---|---|---|---|
| Method | Metric | Group 1 extensor | Group 2 flexor | Group 3 medial | Metric | Group 1 extensor | Group 2 flexor | Group 3 medial |
|
A1 |
DC | 0.94±0.03* | 0.88±0.06* | 0.89±0.05* | DC | 0.95±0.03* | 0.90±0.03* | 0.89±0.06* |
| P | 0.95±0.03 | 0.91±0.08* | 0.87±0.10 | P | 0.92±0.06 | 0.91±0.04* | 0.89±0.07 | |
| R | 0.94±0.05* | 0.86±0.08* | 0.91±0.04* | R | 0.97±0.03 | 0.90±0.07* | 0.90±0.11* | |
| HD(mm) | 12.98±6.44 | 12.84±6.96 | 20.47±10.2 | HD(mm) | 10.51±6.37 | 12.67±3.13 | 31.53±14.24 | |
|
A2 |
DC | 0.84±0.12 | 0.79±0.10 | 0.84±0.10 | DC | 0.89±0.08 | 0.85±0.09 | 0.85±0.09 |
| P | 0.95±0.03 | 0.89±0.09 | 0.89±0.07 | P | 0.92±0.05 | 0.89±0.08 | 0.90±0.09 | |
| R | 0.80±0.20 | 0.77±0.16 | 0.79±0.17 | R | 0.88±0.14 | 0.84±0.13 | 0.83±0.15 | |
| HD(mm) | 20.22±6.07 | 14.75±5.90 | 19.80±9.29 | HD(mm) | 16.70±6.4 | 14.94±5.91 | 31.45±9.93 | |
|
A3 |
DC | 0.86±0.10 | 0.83±0.10 | 0.85±0.07 | DC | 0.89±0.08 | 0.87±0.07 | 0.86±0.07 |
| P | 0.96±0.03 | 0.90±0.09 | 0.89±0.08 | P | 0.92±0.05 | 0.90±0.07 | 0.91±0.07 | |
| R | 0.82±0.18 | 0.82±0.14 | 0.86±0.12 | R | 0.91±0.12 | 0.90±0.06 | 0.87±0.14 | |
| HD(mm) | 15.60±6.71 | 13.42±7.27 | 22.41±9.48 | HD(mm) | 12.85±6.7 | 14.12±4.58 | 32.35±12.31 | |
The average (± SD) values of accuracy measures (Dice’s coefficient (DC), Recall (R), Precision (P), and the Hausdorff distance (HD) for three methods (A1: proposed algorithm, A2: ANTs, and A3: STAPLE) for SCI and ND groups.
* Our proposed algorithm showed equal or higher accuracy than the other two methods.