Table 4.
Obstructive apneas versus central apneas: prediction performance of different models.
| Method | Accuracya (%) | Precisiona (%) | Recalla (%) | F1 scorea (%) |
| Autocorrelation | 88.4 | 81.1 | 53.1 | 64.2 |
| Histogram of movements | 88.5 | 86.7 | 48.2 | 61.9 |
| 2DFFT-CNNb | 89.7 | 69.1 | 75.6 | 72.3 |
| 3D-CNNc | 95.4 | 88.2 | 89.3 | 88.7 |
| 3D-CNN chest and abdomen (annotated) | 90.9 | 71.1 | 81.8 | 76.1 |
| 3D-CNN chest and abdomen (estimated) | 89.3 | 72.1 | 76.0 | 74.0 |
aAccuracy, precision, recall, and F1 score indicate the ratio of correct prediction to the total number of data points, the ratio of correct positive prediction to the total positive prediction, the ratio of correct positive prediction to the total positive data, and the harmonic mean of precision and recall.
b2DFFT-CNN: 2D fast Fourier transform-convolutional neural network
c3D-CNN: 3D convolutional neural network.