Table 7 Output comparison between final model, OR (Original Reader), and OOBTR (Out Of Box Test Reader). Model output compared with OOBTR and OR using multiple evaluation metrics for multiple binary classifications with binary output classes derived from the original multiclass output values
| Â | O.O.B.T.R.- O.R. | Model- O.R. | Model- O.O.B.T.R. |
|---|---|---|---|
Absence or Presence of Opacity | |||
 Precision | 92.50 | 88.64 | 88.89 |
 Recall | 89.53 | 91.19 | 93.62 |
 F1 | 90.95 | 89.89 | 91.12 |
 R-squared | 0.45 | 0.42 | 0.49 |
No Opacity, Mild Opacity VS. Medium & Severe Opacity | |||
 Precision | 77.53 | 81.31 | 82.57 |
 Recall | 75.08 | 84.23 | 84.08 |
 F1 | 76.25 | 82.51 | 83.01 |
 R-squared | 0.21 | 0.44 | 0.46 |
Not Severe or Severe | |||
 Precision | 46.64 | 55.27 | 71.56 |
 Recall | 62.57 | 80.64 | 69.53 |
 F1 | 53.45 | 65.45 | 70.32 |
 R-squared | 0.06 | 0.02 | 0.28 |