Quantification of the performance of the 5 developed neural networks in the developmental phase
| . | NN01 . | NN02 . | NN03 . | NN04 . | NN05 . |
|---|---|---|---|---|---|
| Positive predictive value (%) | 94.0 | 95.6 | 95.8 | 91.3 | 93.6 |
| Negative predictive value (%) | 98.4 | 92.5 | 96.9 | 90.9 | 93.8 |
| Sensitivity (%) | 97.9 | 89.6 | 95.8 | 87.5 | 91.7 |
| Specificity (%) | 95.3 | 96.9 | 96.9 | 93.8 | 95.3 |
| Overall accuracy (%) | 96.4 | 93.8 | 96.4 | 91.1 | 93.8 |
| . | NN01 . | NN02 . | NN03 . | NN04 . | NN05 . |
|---|---|---|---|---|---|
| Positive predictive value (%) | 94.0 | 95.6 | 95.8 | 91.3 | 93.6 |
| Negative predictive value (%) | 98.4 | 92.5 | 96.9 | 90.9 | 93.8 |
| Sensitivity (%) | 97.9 | 89.6 | 95.8 | 87.5 | 91.7 |
| Specificity (%) | 95.3 | 96.9 | 96.9 | 93.8 | 95.3 |
| Overall accuracy (%) | 96.4 | 93.8 | 96.4 | 91.1 | 93.8 |
The positive and negative predictive values, sensitivity and specificity, and the overall accuracy were determined for all 5 neural networks (NN01-NN05) in the developmental cohort during the developmental phase and in a separate clinical validation cohort in the clinical validation phase. Overall, NN01 and NN03 showed the most accurate performance.