Table 2.
Performance of the predictive models in the test cohort.
Radiomics | Clinical | Combined | |
---|---|---|---|
AUC [95% CI] | 0.843 [0.731–0.922] | 0.813 [0.696–0.900] | 0.865 [0.757–0.938] |
Best cut-off value | 0.22 | 0.39 | 0.50 |
Sensitivity (%) | 97.52 | 73.91 | 78.26 |
Specificity (%) | 65.85 | 85.37 | 87.80 |
Negative predictive value (%) | 98.45 | 85.37 | 87.80 |
Positive predictive value (%) | 62.16 | 73.91 | 78.26 |
True positive rate (%) | 99.12 | 73.91 | 78.26 |
False positive rate (%) | 34.15 | 14.63 | 12.20 |
True negative rate (%) | 65.85 | 85.37 | 87.80 |
False negative rate (%) | 1.55 | 26.09 | 21.74 |
False discovery rate (%) | 37.83 | 26.09 | 21.74 |
Accuracy (%) | 78.13 | 81.25 | 84.38 |
Precision (%) | 62.16 | 73.91 | 78.26 |
Youden Index J | 0.6585 | 0.5928 | 0.6607 |
Recall | 0.98 | 0.73 | 0.78 |
P-value | < 0.001 | 0.001 | < 0.0001 |
P < 0.05 indicates that it is statistically significant.
CI confidence interval; AUC area under the curve.