Table 3.
Average volumetric Dice coefficient of our model and for previously published results.
| Brain Stem | Spinal Cord | Esophagus | Pharynx | Larynx | Mandible | Cochlea | |
|---|---|---|---|---|---|---|---|
| Current, DLSu | 0.87 | 0.82 | 0.80 | 0.82 | 0.85 | 0.95 | 0.75 |
| Current, DLSm | 0.87 | 0.82 | 0.82 | 0.82 | 0.85 | 0.95 | 0.76 |
| Fritscher et al. [37] | |||||||
| Ibragimov et al. [38] | 0.87 | ||||||
| Mocnik et al. [39] | |||||||
| Ren X et al. [40] | |||||||
| Zhu et al. [41] | 0.87 | 0.93 | |||||
| Nikolov et al. [36] | 0.84 | 0.88 | 0.94 | 0.70 | |||
| Tong et al. [42] | 0.87 | 0.94 | |||||
| van Rooij et al. [43] | 0.64 | 0.60 | 0.71 | 0.78 | |||
| Rhee et al. [44] | 0.86 | 0.83 | 0.81 | 0.87 | 0.66 | ||
| Liang et al. [34] | 0.90 | 0.88 | 0.87 | 0.91 | 0.82 | ||
| van Dijk et al. [13] | 0.84 | 0.87 | 0.55 | 0.68 | 0.71 | 0.94 | |
| Wong et al. [45] | 0.80–0.83 | 0.79 | |||||
| Zhensong et al. [46] | 0.90 | 0.94 | |||||
| Oktay et al. [47] | 0.79–0.90 | 0.82–0.93 | 0.94–0.99 | ||||
| ParotidG | SMG | Thyroid | Eye | Lens | Optic nerve | Optic chiasm | |
| Current, DLSu | 0.85 | 0.82 | 0.88 | 0.91 | 0.77 | 0.71 | 0.53 |
| Current, DLSm | 0.87 | 0.87 | 0.88 | 0.91 | 0.79 | 0.71 | 0.52 |
| Fritscher et al. [37] | 0.81 | 0.65 | 0.51 | ||||
| Ibragimov et al. [38] | 0.78 | 0.71 | 0.88 | 0.64 | 0.37 | ||
| Mocnik et al. [39] | 0.79 | ||||||
| Ren X et al. [40] | 0.71 | 0.58 | |||||
| Zhu et al. [41] | 0.87 | 0.81 | 0.71 | 0.53 | |||
| Nikolov et al. [36] | 0.86 | 0.77 | 0.95 | 0.80 | 0.70 | ||
| Tong et al. [42] | 0.83 | 0.78 | 0.67 | 0.58 | |||
| van Rooij et al. [43] | 0.83 | 0.82 | |||||
| Rhee et al. [44] | 0.83 | 0.89 | 0.72 | 0.69 | 0.41 | ||
| Liang et al. [34] | 0.85 | 0.84 | 0.69 | ||||
| van Dijk et al. [13] | 0.84 | 0.78 | 0.83 | ||||
| Wong et al. [45] | 0.80 | 0.81–0.82 | 0.85–0.88 | 0.43–0.47 | 0.32–0.38 | ||
| Zhensong et al. [46] | 0.83 | ||||||
| Oktay et al. [47] | 0.83–0.93 | 0.75–0.92 | 0.92–0.97 |
Abbreviations: parotidG, parotid gland; SMG, submandibular gland; DLSu, deep learning-based segmentation using unmatched training set; DLSm, deep learning-based segmentation using matched training set.