Table 2 Performance metrics on test sets, in percentage points, for our method, the proprietary algorithm, and the baselines30,31,32,33.
From: Robust screening of atrial fibrillation with distribution classification
Algorithm | Training | Testing | Acc. | F1 | Sen. | Spe. | AUROC | Noise |
|---|---|---|---|---|---|---|---|---|
Ours | DiagnoStick | DiagnoStick | 96.12 | 66.26 | 71.05 | 97.54 | 97.17 | – |
Ours | SPH | DiagnoStick | 95.76 | 66.29 | 77.63 | 96.79 | 98.20 | – |
Ours | CinC | DiagnoStick | 96.68 | 67.13 | 63.16 | 98.58 | 93.49 | – |
Ours | DiagnoStick | SPH | 93.72 | 80.06 | 75.21 | 97.45 | 97.56 | – |
Ours | SPH | SPH | 95.18 | 85.87 | 87.32 | 96.77 | 98.24 | – |
Ours | CinC | SPH | 92.82 | 77.78 | 74.93 | 96.43 | 96.11 | – |
Ours | DiagnoStick | CinC | 93.22 | 66.67 | 77.54 | 94.73 | 92.44 | – |
Ours | SPH | CinC | 92.72 | 65.88 | 80.43 | 93.89 | 94.48 | – |
Ours | CinC | CinC | 94.30 | 69.80 | 75.36 | 96.11 | 94.85 | – |
Proprietary | – | DiagnoStick | 92.51 | 50.93 | 72.37 | 93.65 | – | – |
Datta et al.30 | \(\hbox {CinC}^\star\) | DiagnoStick | 96.96 | 66.14 | 55.26 | 99.33 | – | 1.27 |
Hong et al.31 | \(\hbox {CinC}^\star\) | DiagnoStick | 94.63 | 0.00 | 0.00 | 100.00 | – | 84.53 |
Zabihi et al.32 | \(\hbox {CinC}^\star\) | DiagnoStick | 96.33 | 54.39 | 40.79 | 99.48 | – | 4.87 |
Mahajan et al.33 | \(\hbox {CinC}^\star\) | DiagnoStick | 95.34 | 37.74 | 26.32 | 99.25 | – | 11.65 |
