Table 1 Performance of our proposed liver segmentation approach compared with state-of-the-art methods. Numbers in parentheses are standard deviations. For the cited studies, scores are reported as presented in the original papers
From: A deep learning framework for automated detection and quantitative assessment of liver trauma
Method | Dataset | Dice (%) | Recall (%) | Precision (%) | RVD (%) | VOE (%) |
|---|---|---|---|---|---|---|
Proposed method | Internal UMHS Dataset (n = 77) | 96.13 (1.49) | 96.00 (2.83) | 96.35 (2.09) | –0.30 (4.24) | 7.40 (2.69) |
Proposed method | 3DIRCAD (n = 20) | 94.64 (2.18) | 95.06 (4.07) | 94.38 (2.75) | 0.83 (5.79) | 10.10 (3.85) |
Ahmad et. al [17] | Subset of 3DIRCAD (n = 5) | 91.83 (1.37) | – | – | 5.59 (6.49) | – |
Lu et. al [18] | 3DIRCAD (n = 20) | – | – | – | 0.97 (3.26) | 9.36 (3.34) |
Christ et. al [19] | 3DIRCAD (n = 20) | 94.3 | – | – | –1.4 | 10.7 |
Lebre et. al [22] | 3DIRCAD (n = 20) | 88 (3) | 87(5) | 89 (4) | – | – |
Kavur et. al [39] | Subset of 3DIRCAD (n = 10) | 92.0 | – | – | 6.42 | – |
Xi et. al [40] | LiTS (n = 70) | 94.9 | – | – | 2.1 | 9.5 |