Table 3.
Comparison of the quantified parameters for each type of segmentation.
| Parameters | Sample ID | WF | HQ | ML |
|---|---|---|---|---|
| DR [mm/year] | 1 | 0.252 (−3%) | 0.261 | 0.243 (−7%) |
| 2 | 0.205 (−2%) | 0.209 | 0.208 (−) | |
| 3 | 0.417 (−10%) | 0.462 | 0.436 (−6%) | |
| BIC [%] | 1 | 62.94 (−) | 62.97 | 70.44 (+12%) |
| 2 | 81.07 (+1%) | 80.22 | 80.30 (−) | |
| 3 | 60.13 (+33%) | 45.14 | 51.61 (+14%) | |
| BV/TV [%] | 1 | 47.88 (−1%) | 48.14 | 48.45 (+1%) |
| 2 | 55.23 (+1%) | 54.93 | 54.67 (−) | |
| 3 | 41.25 (+4%) | 39.64 | 41.25 (+4%) |
We consider high quality (HQ) segmentation as the reference (in bold). Percentage values in brackets represent the relative differences between workflow (WF) and machine learning (ML) segmentation compared to the HQ segmentation. (−) means that the difference was less than 1%.