Table 3.
Effectiveness of the three models in discriminating AIS-MIA from IAC in the derivation, internal validation, and external validation sets.
| Az (95% CI) | Cutoff value | SEN (%) | SPE (%) | PPV (%) | NPV (%) | ACC (%) | F1-score | MCC | Model-fitting information | |||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| AIC (%) | R2 value | |||||||||||
| Derivation set | ||||||||||||
| Clinical-radiographic model | 0.779 (0.747-0.809) | >0.471 | 64.8 | 76.8 | 71.9 | 70.3 | 71.0 | 0.682 | 0.443 | 47.5 | 0.225 | |
| Radiomics model | 0.865 (0.839-0.889) | >0.456 | 75.3 | 81.5 | 78.9 | 78.2 | 78.5 | 0.770 | 0.569 | 63.2 | 0.398 | |
| Combined model | 0.872 (0.846-0.895) | >0.494 | 78.7 | 82.0 | 80.1 | 80.7 | 80.4 | 0.794 | 0.607 | 64.5 | 0.415 | |
| Internal validation set | ||||||||||||
| Clinical-radiographic model | 0.810 (0.762-0.852) | >0.416 | 66.9 | 76.8 | 72.7 | 71.6 | 72.1 | 0.697 | 0.439 | 53.0 | 0.279 | |
| Radiomics model | 0.884 (0.843-0.917) | >0.531 | 78.1 | 83.5 | 81.4 | 80.6 | 81.0 | 0.797 | 0.618 | 66.8 | 0.444 | |
| Combined model | 0.887 (0.847-0.920) | >0.560 | 80.1 | 84.8 | 82.9 | 82.2 | 82.5 | 0.815 | 0.650 | 68.4 | 0.466 | |
| External validation set | ||||||||||||
| Clinical-radiographic model | 0.855 (0.799-0.901) | >0.419 | 71.3 | 80.8 | 77.0 | 75.7 | 76.3 | 0.740 | 0.523 | 60.9 | 0.367 | |
| Radiomics model | 0.924 (0.878-0.957) | >0.444 | 83.5 | 84.2 | 83.5 | 84.2 | 83.8 | 0.835 | 0.613 | 73.3 | 0.535 | |
| Combined model | 0.917 (0.869-0.951) | >0.472 | 84.5 | 86.1 | 85.4 | 85.3 | 85.4 | 0.850 | 0.707 | 73.8 | 0.542 | |
Az, area under the receiver operating curve; CI, confidence interval; SEN, sensitivity; SPE, specificity; PPV, positive predictive value; NPV, negative predictive value; ACC, accuracy; MCC, Matthews correlation coefficient; AIC, Akaike information criterion; AIS, adenocarcinoma in situ; MIA, minimally invasive adenocarcinoma; IAC, invasive adenocarcinoma.