Table 3.
Comparison of diagnostic performance between multiparametric MR radiomics models and other approaches in the training and validation sets
| Training Set | External Validation Set | |||||||
|---|---|---|---|---|---|---|---|---|
| Comparison | AUC | P* | Sensitivity | Specificity | AUC | Sensitivity | Specificity | |
| Combined radiomics model | Multiparametric MR (conventional + diffusion + perfusion MR) |
0.90 (0.82, 0.98) |
91.4% | 76.9% | 0.85 (0.71, 0.99) |
71.4% | 90.0% | |
| Single radiomics model | Conventional MR | 0.76 (0.63, 0.88) |
0.012 | 51.4% | 88.5% | 0.74 (0.67, 0.97) |
78.6% | 75.0% |
| Diffusion MR | 0.78 (0.66, 0.90) |
0.014 | 77.1% | 76.9% | 0.53 (0.33, 0.73) |
100% | 20% | |
| Perfusion MR | 0.88 (0.80, 0.97) |
0.427 | 65.7% | 96.2% | 0.71 (0.52, 0.89) |
85.7% | 60.0% | |
| Single parameter | Mean ADC | 0.57 (0.42, 0.73) |
<0.001 | 77.1% | 46.2% | 0.57 (0.38, 0.77) |
78.6% | 45.0% |
| Minimum ADC | 0.61 (0.46, 0.76) |
<0.001 | 71.4% | 57.7% | 0.50 (0.30, 0.70) |
50.0% | 65.0% | |
| Mean CBV | 0.77 (0.64, 0.87) |
<0.001 | 65.7% | 84.6% | 0.58 (0.40, 0.75) |
100.0% | 30.0% | |
| Maximum CBV | 0.79 (0.66, 0.88) |
<0.001 | 62.9% | 92.3% | 0.58 (0.40, 0.75) |
78.6% | 45.0% | |
Note: Numbers in parentheses are 95% confidence intervals.
*P-value refers to the significance among the differences of the AUCs between the multiparametric MR radiomics model and the other model. The Bonferroni-corrected significance level of P < 0.017 was used when comparing between a multiparametric MR radiomics model and 3 single radiomics models, and P < 0.0125 between a multiparametric radiomics model and 4 single parameter approaches.