Table 3 Comparison of the segmentation accuracy of the pipeline before and after correction by the Random Forest (RF) models
From: A deep learning pipeline for automatic analysis of multi-scan cardiovascular magnetic resonance
CMR sequence | Anatomical structure | Dice | |
|---|---|---|---|
Before RF | After RF | ||
LGE | LV cavity | 0.86 (0.06) | 0.90 (0.06) |
LV myocardium | 0.82 (0.05) | 0.83 (0.04) | |
Scar | 0.54 (0.19) | 0.57 (0.18) | |
Native T1 | LV cavity | 0.81 (0.08) | 0.94 (0.05) |
LV myocardium | 0.87 (0.08) | 0.88 (0.08) | |
Post-contrast T1 | LV cavity | 0.85 (0.06) | 0.96 (0.04) |
LV myocardium | 0.86 (0.05) | 0.92 (0.03) | |
Native T2 | LV cavity | 0.87 (0.04) | 0.96 (0.03) |
LV myocardium | 0.90 (0.04) | 0.91 (0.05) | |