Table 2.
Performance of pixelwise validation.
| Model | Metrics | Subject 1 | Subject 2 | Subject 3 |
|---|---|---|---|---|
| Individual Models | Sensitivity | 0.87 | 0.82 | 0.89 |
| Precision | 0.92 | 0.85 | 0.89 | |
| F1 | 0.89 | 0.84 | 0.89 | |
| Combined Model | Sensitivity | 0.80 | 0.79 | 0.91 |
| Precision | 0.93 | 0.86 | 0.88 | |
| F1 | 0.86 | 0.82 | 0.90 |
Precision, sensitivity, and F1-score of 2048 × 2048 pixel image for each subject with individual trained UNetDense models and one combined UNetDense model. Pixel evaluation is based on the confusion matrix of Table 1.