Table 2.
Logistic regression model results
| OR | 10th, 90th OR | AUC | 10th, 90th AUC | |
|---|---|---|---|---|
| DVH models | 0.66 ± 1.1 | 0.52, 0.80 | ||
| V40 | 5.58 | 0, 10.77 | ||
| V45 | 2.16 | 0, 8.89 | ||
| DVHEQD2 models | 0.66 ± 0.11 | 0.51, 0.80 | ||
| V40 | 4.90 | 0, 12.10 | ||
| V35 | 1.34 | 0, 9.30 | ||
| DO models | 0.70 ± 0.11 | 0.55, 0.85 | ||
| NTGDM busyness | − 0.07 | − 0.10, − 0.04 | ||
| 90 percentile | 0.01 | 0, 0.02 | ||
| GLRLM LongRunGrayLevelEmphasis | 0.51a | 0, 2.05a | ||
| DOEQD2 models | 0.71 ± 0.11 | 0.57, 0.84 | ||
| NTGDM busyness | − 0.07 | − 0.11, − 0.04 | ||
| 90 percentile | 0.02 | 0, 0.11 | ||
| GLSZM LowGrayLevelZoneEmphasis | 25.96 | − 117.66, 0 |
Features corresponding to the model showed only features that were selected more than 50% of the time. OR was adjusted to the actual value of features (not normalized value)
aReported normalized value due to very low OR