Table 2. Testing scores using nested leave-one-patient-out cross validation on all classification tasks over five patients, using features of SO-BSS, AFFTr2DF and NDI.
Bold fonts mark the highest scores and the corresponding features.
| Task | Feature (with signal used) |
Accuracy (mean±s.d.%) |
Precision (mean±s.d.%) |
Recall (mean±s.d.%) |
|---|---|---|---|---|
| FS presence | SO-BSS (BSPM) | 90.6±1.0 | 96.8±0.7 | 89.7±1.2 |
| SO-BSS (ECG) | 90.6±0.6 | 97.2±0.7 | 89.4±1.2 | |
| AFFTr2DF (V1) | 56.8±2.2 | 73.2±1.9 | 61.3±2.7 | |
| NDI (BSPM) | 60.8±2.3 | 75.9±1.7 | 65.2±3.6 | |
| AF sustainability | SO-BSS (BSPM) | 93.1±0.6 | 90.4±3.0 | 93.0±3.1 |
| SO-BSS (ECG) | 94.2±1.2 | 92.4±2.6 | 93.4±3.3 | |
| AFFTr2DF (V1) | 58.5±2.0 | 48.2±3.7 | 54.1±4.0 | |
| NDI (BSPM) | 61.6±2.3 | 52.1±5.1 | 57.4±5.7 | |
| AF sustainability (evaluated on FS-driven episodes) |
SO-BSS (BSPM) | 91.5±1.2 | 69.6±6.8 | 83.0±11.6 |
| SO-BSS (ECG) | 92.7±1.8 | 74.7±6.3 | 82.4±13.1 | |
| AFFTr2DF (V1) | 60.3±3.0 | 20.5±2.6 | 55.7±4.8 | |
| NDI (BSPM) | 63.2±3.0 | 22.6±4.2 | 56.5±4.6 | |
| FS location | SO-BSS (BSPM) | 80.0±6.6 | 81.2±9.0 | 75.0±8.4 |
| SO-BSS (ECG) | 61.0±5.2 | 59.3±6.8 | 53.6±8.4 |