Table 1.
Characteristics of included studies: articles published using AI and/or ML to detect IUGR during various pregnancy timings.
| Author | Study Type | Setting | Sample (n) | Preg Time (weeks) | Methods | AI/ML Model | Outcomes | Measures (Accuracy) |
|---|---|---|---|---|---|---|---|---|
| Guo, Z. [50] | Obs. retr. case-control | China | 2199 | 12–28 | DNA profiling | LR | Using ML to predict FGR and BW | 79% |
| Dahdoud, S. [58] | Obs. retr. case-control | USA | 80 | 18–39 | MRI | RUSBoost | Using ML to predict FGR and BW | 86% |
| Lunghi, F. [45] | Obs. retr. case-control | Italy | 909 | 30–35 | FHR by CTG | SVM | Realizing an automatic system for identified FGR | 84% |
| Magenes, G. [48] | Obs. retr. case-control | Italy | 100 | 30–35 | FHR by CTG | SVM | Realizing an automatic system for identified FGR | 78% |
| Signorini, M. [60] | Obs. retr. case-control | Italy | 120 | 30–35 | FHR by CTG | RF (best) | Find the best classification ML model for identifying IUGR | 91% |
| Crockart, I.C. [27] | Obs. prosp. case-control | USA and S. Africa | 6004 | 20–29 | FHR by CTG | Stochastic Gradient Descent, LR & RF | Using ML to predict FGR and BW | 91% |
| Bahado–Singh, R. [46] | Obs. retr. case-control | USA | 80 | Delivery | Biochemical | SVM | Find the best classification ML model for identifying IUGR | 80% |
| Pini, N. [47] | Obs. retr. case-control | Italy | 262 | 36–37 | FHR by CTG | RBF-SVM | Build a ML screener for late IUGR | 93% |
| Magenes, G. [51] | Obs. retr. case-control | Italy | 122 | 30–35 | FHR by CTG | RF & LR | Find the best classification ML model for identifying IUGR | RF = 85%; LR = 83% |
| Xu, C. [52] | Obs. retr. nested case-control | China | 810 | 12–27 | DNA profiling | SVM & LR | Find the best classification ML model for identifying IUGR | 83% |
| Buscema, M. [54] | Obs. retr. case-control | Italy | 46 | Delivery | Biochemical | ACM & ACS | Find the best classification ML model for identifying IUGR | 87% |
| Foltran, F. [55] | Obs. prosp. case-control | Italy | 46 | 20–32 | Biochemical | BN | Realizing an automatic system for identified FGR | 90% |
| Street, M.E. [56] | Obs. retr. case-control | Italy | 48 | 20–32 | Biochemical | ANNS | Find the best classification ML model for identifying IUGR | 89% |
| Ferrario, M. [57] | Obs. retr. case-control | Italy | 59 | 27–34 | FHR by CTG | LZ complexity | Realizing an automatic system for identified FGR | 91% |
| Deval, R. [49] | Obs. retr. case-control | India | 214 | - | Biochemical | SVM, MLP | Using ML models to predict IUGR | SMO = 95.5%; MLP = 8.5% |
| Gómez–Jemes, L. [53] | Obs. retr. case-control | Slovenia | 95 | 24–38 | Doppler indices: UA, sFIt-1, and PIGF values | Multi-models (extra-trees, RF) | Using ML models to predict pre-Eclampsia, IUGR | Extra trees = 78%, RF = 73% |
| Sufriyana, H. [59] | Obs. prosp. case-control | Slovenia | 95 | 24–37 | Doppler indices: UA, sFIt–1, and PIGF values | CVR | Using ML models to predict pre-Eclampsia, IUGR | CVR = 90.6% |
| Aslam, N. [61] | Obs. retr. case-control | Italy | 382 | 30–37 | FHR by CTG | SVM & RF | Using ML models to predict IUGR | RF = 97% |
| Gürgen, F. [62] | Obs. retr. case-control | Turkey | 44 | <38 | Doppler indices: PI & RI of UA, MCA, DV, and AFI | SVM | Using ML models to predict IUGR | SVM = 81% |
| Van, S.N. [25] | Obs. prosp. case-control | China | 75 | - | Physiological, clinical, and socioeconomic | Seven ML algorithms | Identify the latent risk clinical attributes using the ML algorithms. | 94.73% |