Table 1.
Summary of previous studies for the prediction of preterm birth by obtaining linear, time-frequency, and nonlinear EHG features.
| Authors | Classifier type | Dataset | Length | Features | Accuracy | Sensitivity | Specificity | AUC |
|---|---|---|---|---|---|---|---|---|
| Acharya et al., (2017) (6) | Support vector machine (SVM) Radial Basis Function (RBF) |
TPEHG DB (34 recordings of preterm and 262 of the term) |
Complete recording | Empirical Mode Decomposition (EMD) Wavelet Packet Decomposition (WPD) Entropy-based methods |
96.25 | 95.08 | 97.33 | 0.96 |
| Jager et al., (2018) (7) | Quadratic Discriminant | TPEHG DB TPEHG DS |
Complete recording | Sample Entropy Medium Frequency Maximum Frequency |
100.00 | 100.00 | 100.00 | 1.00 |
| Nieto del Amor et al., (2021) (9) | Linear Discriminant | TPEHG DB | 120-seconds window Complete recording |
Dominant frequency Normalized Energy Power spectrum deciles (D3, D6, D8, D9) Entropy-based methods |
89.2% ± 2.4 | 98.4% ± 1.9 | 79.9% ± 4.9 | 0.93 |
| Hoseinzadeh & Amirani (2018) (10) | RBF SVM | TPEHG DB | N/A | EMD WPD Feature extraction by autoregressive models |
97.1 | 95 | 99 | N/A |
N/A, Not available.