Table 3.
Performance of the machine-learning algorithm.
| Comparisons | Speech-task | Assoc. criterion | Youden index | Se (%) | Sp (%) | PPV (%) | NPV (%) | Acc (%) | AUC |
|---|---|---|---|---|---|---|---|---|---|
| PWS vs. C | Vowel | −0.24 | 0.75 | 88.7 | 86.3 | 90.0 | 84.6 | 87.7 | 0.934 |
| S1 | 0.12 | 0.66 | 85.7 | 80.8 | 85.7 | 80.8 | 83.6 | 0.906 | |
| S2 | −0.09 | 0.62 | 81.3 | 80.9 | 87.1 | 73.1 | 81.1 | 0.881 | |
| cC vs. cPWS | Vowel | 0.23 | 0.94 | 93.8 | 100 | 100 | 90.0 | 96.0 | 0.993 |
| S1 | 0.06 | 0.85 | 90.6 | 94.4 | 96.7 | 85.0 | 92.0 | 0.970 | |
| yC vs. yPWS | Vowel | 0.66 | 0.80 | 92.3 | 87.9 | 90.0 | 90.6 | 90.3 | 0.975 |
| S1 | 0.07 | 0.81 | 88.4 | 93.1 | 95.0 | 84.4 | 90.3 | 0.973 | |
| cC vs. yC | Vowel | 0.39 | 0.88 | 93.3 | 95.5 | 93.3 | 95.5 | 94.3 | 0.991 |
| S1 | 0.43 | 0.94 | 96.7 | 97.5 | 96.7 | 97.5 | 97.1 | 0.998 | |
| cPWS vs. yPWS | Vowel | 0.20 | 0.86 | 94.4 | 91.2 | 85.0 | 96.9 | 92.3 | 0.956 |
| S1 | 0.24 | 0.89 | 100 | 88.9 | 80.0 | 100 | 92.3 | 0.933 |
C, the whole group of controls; cC, controls in prepubertal age; yC, younger adult controls in postpubertal age; PWS, the whole group of people with stuttering; cPWS, children people with stuttering in the prepubertal age; yPW, younger adult people with stuttering in the postpubertal age; Acc, accuracy; AUC, area under the curve; NPV, negative predictive value; PPV, positive predictive value; Se, sensitivity; Select Sent., selective sentence which consists of the emission of the sentence of the connected speech supposed to induce stuttering in patients; Sp, specificity; Stand. Sent., standard sentence which is the sentence of the connected speech used in our previous papers in the field of voice analysis. Instances refer to the number of subjects considered in each comparison. Cross-validation refers to standardized procedures of a machine-learning algorithm (see the text for details).