Table 3.
Prediction accuracy of the neural network algorithm using the neuropsychological screening test dataset
| Prediction | Number of subjects | Accuracy of 10 trials (mean ± SD%) |
SE(%) | SP(%) | PPV(%) | NPV(%) | AUC | |
|---|---|---|---|---|---|---|---|---|
| Balanced dataset | CI vs NC | 3231: 3217 | 96.66 ± 0.52 | 96.0 | 96.8 | 97.0 | 95.8 | 0.964 |
| MCI vs NC | 3217: 3217 | 96.60 ± 0.45 | 96.0 | 97.4 | 97.6 | 95.6 | 0.967 | |
| ADD vs MCI vs NC | 3235: 3217: 3217 | 95.49 ± 0.53 | ||||||
| Clinic-based dataset | CI vs NC | 11,709: 3217 | 97.23 ± 0.32 | 97.4 | 95.2 | 98.6 | 91.3 | 0.963 |
| MCI vs NC | 6002: 3217 | 97.05 ± 0.38 | 97.5 | 96.4 | 98.1 | 94.8 | 0.968 | |
| ADD vs MCI vs NC | 5707: 6002: 3217 | 96.34 ± 1.03 |
SD Standard deviation, SE Sensitivity, SP Specificity, PPV Positive predictive value, NPV Negative predictive value, AUC Area under the curve, CI Cognitive impairment, NC Normal cognition, MCI Mild cognitive impairment