TABLE III.
diagnostic ability of AHICNN for the ahi cutoffs= 1e/h, 5 e/h, and 10 e/h in the CHAT, UofC and BUH test databases
| Estimated AHI |
CHAT test set | UofC test set | BUH test set | ||||||
|---|---|---|---|---|---|---|---|---|---|
|
| |||||||||
| AHI = 1 e/h |
AHI =5 e/h |
AHI = 10 e/h |
AHI = 1 e/h |
AHI =5 e/h |
AHI = 10 e/h |
AHI = 1 e/h |
AHI =5 e/h |
AHI = 10 e/h |
|
| Se (%) | 71.2 | 83.7 | 83.9 | 90.8 | 76.0 | 79.5 | 88.8 | 61.1 | 65.0 |
| Sp (%) | 81.8 | 100 | 99.3 | 36.4 | 88.6 | 95.8 | 53.2 | 93.7 | 96.9 |
| PPV (%) | 72.4 | 100 | 92.9 | 85.4 | 79.8 | 83.5 | 83.8 | 81.5 | 81.3 |
| NPV (%) | 81.0 | 97.0 | 98.2 | 49.1 | 86.2 | 94.6 | 63.5 | 84.2 | 93.0 |
| LR+ | 3.92 | N.D | 117.84 | 1.43 | 6.68 | 18.90 | 1.90 | 9.72 | 20.69 |
| LR− | 0.35 | 0.16 | 0.16 | 0.25 | 0.27 | 0.21 | 0.21 | 0.42 | 0.36 |
| Acc (%) | 77.6 | 97.4 | 97.8 | 80.1 | 83.9 | 92.3 | 79.2 | 83.5 | 91.3 |
| kappa | 0.515 | 0.422 | 0.423 | ||||||
CNN = Convolutional neural network, AHI = apnea-hypopnea index, Se = sensitivity, Sp = specificity, PPV = positive predictive value, NPV = negative predictive value, LR+ = positive likelihood ratio, LR- = negative likelihood ratio, Acc = accuracy, kappa = Cohen’s kappa index, N.D = not defined, CHAT = Childhood Adenotonsillectomy Trial, UofC = University of Chicago, BUH = Burgos University Hospital.