Table 5.
ML applications in PTB research.
| Reference | ML application | Input | ML technique | Main output |
|---|---|---|---|---|
| (86) | Complication prediction | Clinical and biochemical factors | DT, SVM and RF | AUC = 68% and Ac = 81% for spontaneous and provider-initiated cases |
| (87) | Complication prediction | Clinical and biochemical factors | ANN | AUC = 79.8%, Se = 62.7%, Sp = 84.6%, PPV = 23.2% and NPV = 97.0% |
| (88) | Complication prediction | Electronic health records | LSTM | AUC = 0.744, Se = 0.682, Sp = 0.743 and PPV = 0.028 for extreme cases |
| (89) | Complication prediction | Electronic health records | LSTM | AUC = 0.651, Ac = 0.739, Se = 0.407 and Sp = 0.982 |
| (90) | Biomarker discovery | Biochemical markers | RF | PGA2, 15DO12,14-PGJ2, BCPGE2, 13,14DHK-PGF2a, RVD1, LTE4, LTB4, linolenic acid and IL-10 as novel biomarkers |
| (91) | Biomarker discovery | Metabolomics data | RF | FA(17:1), FA(24:6), FA(14:2), CAR(18:2), hexanoylcarnitine, FA(14:0(Ke)), FA(26:1), raffinose, PC(18:0/16:3), FA(16:3), glycocholic acid, PC(33:4), FA(22:5), FA(14:1(Ke)), heptadecanoic acid, FA(19:1) and FA(14:1) as novel biomarkers |
| (92) | Biomarker discovery | Metabolomics, proteomics and transcriptomics data | RF | IL-6, IL-1RA, G-CSF, RARRES2, CCL3, ANGPTL4, PAD12, TfR, and metabolites from glutamine/glutamate metabolism, and valine/leucine/isoleucine biosynthesis pathways as novel biomarkers |
| (93) | Complication prediction | Electrohysterography recordings | RF | AUC = 0.999, Ac = 99.23%, Se = 98.40%, Sp = 99.76% and Pr = 95.86% |
| (94) | Complication prediction | Clinical parameters | KNN; RF | AUC = 1.00, Ac = 0.95, Se = 0.67, Sp = 1.00, G-means = 0.82 for PTB and the potential value of performing cervical cerclage to prolong the pregnancy; MAE = 3.521, MSE = 4.560 and R = 0.752 for timing of spontaneous delivery |
ML, machine learning; PTB, preterm birth; DT, decision tree; SVM, support vector machines; RF, random forest; ANN, artificial neural networks; LSTM, long short-term memory; KNN, k-nearest neighbors; AUC, area under the receiver operating characteristic curve; Ac, accuracy; Se, sensitivity; Sp, specificity; PPV, positive predictive value; NPV, negative predictive value; Pr, precision; MAE, mean absolute error; MSE, mean squared error.