Statistical limits in sonographic estimation of birth weight

Machine learning approach for the prediction of macrosomia

Fetal macrosomia is associated with maternal and newborn complications due to incorrect fetal weight estimation or inappropriate choice of delivery models. The early screening and evaluation of macrosomia in the third trimester can improve delivery outcomes and reduce complications. However, traditional clinical and ultrasound examinations face difficulties in obtaining accurate fetal measurements during the third trimester of pregnancy. This study aims to develop a comprehensive predictive model for detecting macrosomia using machine learning (ML) algorithms. The accuracy of macrosomia prediction using logistic regression, k-nearest neighbors, support vector machine, random forest (RF), XGBoost, and LightGBM algorithms was explored. Each approach was trained and validated using data from 3244 pregnant women at a hospital in southern China. The information gain method was employed to identify deterministic features associated with the occurrence of macrosomia. The performance of six ML algorithms based on the recall and area under the curve evaluation metrics were compared. To develop an efficient prediction model, two sets of experiments based on ultrasound examination records within 1-7 days and 8-14 days prior to delivery were conducted. The ensemble model, comprising the RF, XGBoost, and LightGBM algorithms, showed encouraging results. For each experimental group, the proposed ensemble model outperformed other ML approaches and the traditional Hadlock formula. The experimental results indicate that, with the most risk-relevant features, the ML algorithms presented in this study can predict macrosomia and assist obstetricians in selecting more appropriate delivery models.

Validity of two-dimensional ultrasound for determining extreme foetal weights to term

The aim of this study was to assess the validity of ultrasound carried out within seven days prior to birth at term for the calculation of full term estimated foetal weight (EFW) in order to diagnose extreme foetal weight, performed using a single-center retrospective cohort study of 2500 pregnant women. Ultrasound calculations of EFW with a standard error of less than 10% showed an accuracy ratio of 75.1% for the total sample. This percentage was similar for appropriate for gestational age (78.6%) and normal foetal weights (77.5%) but decreased significantly (p < .01) in the case of extreme foetal weights. The simple error increased in both the high and low EFWs, taking on positive values for the low weights and negative values for the high weights. As for the percentage error values, there was a tendency for positive errors for low weights and negative errors for high weights; this led to a tendency to overestimate low foetal weights and underestimate high foetal weights.IMPACT STATEMENTWhat is already known on this subject? Two-dimensional ultrasound is currently the principal tool used in obstetrics to evaluate foetal growth, mainly through the calculation of EFW. Foetal weight represents an important prognostic factor in perinatal results, with a greater risk of adverse effects in cases of extreme foetal weights. In this sense, there are few studies that assess the validity of EFW calculations focussing on extreme foetal weights to term.What do the results of this study add? The ultrasound estimates of EFW with an error lower than 10% in the seven days prior to birth showed an accuracy ratio of 75.1% for the total sample. This percentage was similar to appropriate for gestational age weights (78.6%) and of normal weights (77.5%), but decreased significantly (p < .01) in the case of extreme foetal weights: small for gestational age (52.1%), large for gestational age (68.2%), microsomia (49.1%), and macrosomia (61%). Likewise, we found high specificity and low sensitivity for ultrasound diagnosis of extreme foetal weights.What are the implications of these findings for clinical practice and/or future research? The validity of ultrasound EFW is influenced by extreme foetal weights, with a tendency to overestimate low weights and underestimate high weights, which represents a clinically important finding.

Prediction of fetal weight at varying gestational age in the absence of ultrasound examination using ensemble learning

Obstetric ultrasound examination of physiological parameters has been mainly used to estimate the fetal weight during pregnancy and baby weight before labour to monitor fetal growth and reduce prenatal morbidity and mortality. However, the problem is that ultrasound estimation of fetal weight is subject to population's difference, strict operating requirements for sonographers, and poor access to ultrasound in low-resource areas. Inaccurate estimations may lead to negative perinatal outcomes. This study aims to predict fetal weight at varying gestational age in the absence of ultrasound examination within a certain accuracy. We consider that machine learning can provide an accurate estimation for obstetricians alongside traditional clinical practices, as well as an efficient and effective support tool for pregnant women for self-monitoring. We present a robust methodology using a data set comprising 4212 intrapartum recordings. The cubic spline function is used to fit the curves of several key characteristics that are extracted from ultrasound reports. A number of simple and powerful machine learning algorithms are trained, and their performance is evaluated with real test data. We also propose a novel evaluation performance index called the intersection-over-union (loU) for our study. The results are encouraging using an ensemble model consisting of Random Forest, XGBoost, and LightGBM algorithms. The experimental results show the loU between predicted range of fetal weight at any gestational age that is given by the ensemble model and ultrasound respectively. The machine learning based approach applied in our study is able to predict, with a high accuracy, fetal weight at varying gestational age in the absence of ultrasound examination.

[Breech Presentation: CNGOF Guidelines for Clinical Practice - Case Selection for Trial of Labour]

OBJECTIVE

The objective of this chapter is to examine on the basis of the knowledge currently available the criteria available before labour for selecting women who would be eligible for trial of vaginal delivery.

METHODOLOGY

Bibliographical research in French and English using the Medline and Cochrane databases between 1980 and 2019 and the recommendations of international societies.

RESULTS

It is recommended to offer women who wish to attempt a vaginal delivery at term a pelvimetry to decide with them on their mode of delivery (Grade C). The pelvimetric standards used at the time of the PREMODA study were anteroposterior diameter of inlet≥105mm, a transverse diameter of inlet≥120mm, a transverse interspinous diameter≥100mm. However, since there is no evidence about which pelvic measures to use, nor any evidence to set decision-making thresholds other than those set in published studies, the selected decision-making thresholds can be adjusted according to gestational age at delivery or fetal biometrics (Professional consensus). There is no argument for recommending the practice of pelvimetry in the case of delivery before 37 weeks gestational age (Professional consensus) and in the case of breech presentation discovered at the time of beginning of labour, the absence of pelvimetry alone does not contraindicate the attempt of vaginal delivery (Professional consensus). There is insufficient data to recommend the systematic use of fetal weight estimation and/or biparietal diameter measurement as acceptance criteria for a vaginal delivery attempt. In the event of a known fetal weight estimation before birth greater than 3800g, a cesarean section is to be preferred (Professional consensus). The breech presentation is not in itself a contraindication to an attempt of vaginal delivery for a small fetus for gestational age (Professional consensus). The presentation of the non-frank breech is not in itself a contraindication to an attempt of vaginal delivery (Professional consensus). In the case of premature breech delivery, current data do not allow to recommend one delivery route over another (Professional consensus). It is recommended to check the absence of hyperextension of the fetal head by ultrasound before an attempt of vaginal delivery (Professional consensus) and to prefer a cesarean section if such a position is found (Professional consensus). It is not recommended to propose a caesarean section with the sole reason of nulliparity (Grade C). The history of cesarean section is not in itself a contraindication to an attempt of vaginal delivery in the case of fetal breech presentation (Professional consensus). Premature rupture of the membranes is not in itself a contraindication to an attempt of vaginal delivery (Professional consensus).

CONCLUSION

A number of the factors analyzed in this chapter are to be incorporated into the decision-making process in order to choose with the woman whose fetus is in breech presentation the delivery route.

Different formulas, different thresholds and different performance-the prediction of macrosomia by ultrasound

OBJECTIVE

The sonographic prediction of fetal macrosomia affects obstetrical decision regarding the timing and mode of delivery. We aimed to compare the accuracy of various formulas for prediction of macrosomia at different thresholds.

STUDY DESIGN

This was a retrospective cohort study of singleton gestations at term, with fetal biometrical measurements taken up to 7 days prior to delivery (2007 to 2014). Sonographic estimated fetal weight was calculated using 20 previously published formulas. Macrosomia prediction was evaluated for every formula utilizing: (1) measures of accuracy (sensitivity, specificity and so on); (2) comparison of the systematic and random errors (SE and RE), and the proportion of estimates within 10% of actual birth weight for macrosomic and non-macrosomic neonates. Performance measurements were evaluated for different macrosomia thresholds: 4000, 4250 and 4500 g. Best performing formula for every threshold was defined as the one with the lowest Euclidean distance (=SQRT(SE²+RE²)).

RESULTS

Out of 7977 women who met the inclusion criteria, 754 (9.4%) delivered a neonate weighing ⩾4000 g, 266 (3.3%) delivered a neonate weighing⩾4250 g and 75 (0.9%) delivered a neonate weighing⩾4500 g. Considerable variability was noted between the accuracy parameters of the different formulas, with Woo's formula integrating Abdominal circumference (AC) and femur length (FL) as the most sensitive formula with the highest negative predictive value for all thresholds and Woo's formula using AC, FL and biparietal diameter (BPD) as the most specific for all thresholds. The same formula also demonstrated the best overall accuracy. Regardless of threshold chosen, 80% or more of formulas demonstrated negative systematic error, meaning lower EFW than actual birthweight. As for the Euclidean distance, Hadlock's formula (AC, FL and BPD) ranked the highest for the 4000 and 4250 g thresholds, whereas Shepard's formula (AC and BPD) ranked the highest for the 4500 g threshold.

CONCLUSION

Considerable variability exist between formulas for prediction of neonatal macrosomia. Formulas by Hadlock's and Shepard's utilizing AC, BPD±FL were most accurate for macrosomia prediction at 4000, 4250 and 4500 g thresholds, respectively.

Risk factors for emergency caesarean section in planned vaginal breech delivery

PURPOSE

To identify risk factors for emergency caesarean section in women attempting a vaginal breech delivery at term.

METHODS

Data from 1092 breech deliveries performed between 1998 and 2013 at a Swiss cantonal hospital were extracted from an electronic database. Of the 866 women with a singleton, full term pregnancy, 464 planned a vaginal breech delivery. Fifty-seven percent (265/464) were successful in delivering vaginally. Multivariate regression analyses of risk factors were performed, and neonatal and maternal complications were compared.

RESULTS

Risk factors for failed vaginal delivery were peridural anaesthesia (OR 2.05; 95 % CI 1.09-3.84; p = 0.025), nulliparity (OR 2.82; 95 % CI 1.87-4.25; p < 0.001), high birth weight (OR 1.17; 95 % CI 1.04-1.30; p = 0.006) and induction of labour (OR 1.56; 95 % CI 1.003-2.44; p = 0.048). Maternal age, height and weight; gestational age; or newborn length and head circumference were not associated with an unplanned caesarean section. The rate of successful vaginal delivery in the low risk sub-group (multiparous women without induction of labour) was 58-83 %, depending on birth weight category. The likelihood of success for the high risk sub-group (nulliparous women with induction of labour) fell below a third at neonatal birth weights >3250 g. Complication rates were low in the cohort.

CONCLUSIONS

Use of peridural anaesthesia, nulliparity, high birth weight and induction of labour were risk factors for unsuccessful vaginal breech delivery requiring an unplanned caesarean section. Awareness of these risk factors is useful when counselling women who are considering a vaginal breech delivery.

Intrapartum sonographic weight estimation

PURPOSE

To evaluate the accuracy of intrapartum sonographic weight estimation (WE).

MATERIALS AND METHODS

This retrospective, cross-sectional study included 1958 singleton pregnancies. Inclusion criteria were singleton pregnancy with cephalic presentation, vaginal delivery and ultrasound examination with complete biometric parameters performed on the day of delivery during the latent or active phase of labor, and absence of chromosomal or structural anomalies. The accuracy of intrapartum WE was compared to a control group of fetuses delivered by primary cesarean section at our perinatal center and an ultrasound examination with complete biometric parameters performed within 3 days before delivery (n = 392). Otherwise, the same inclusion criteria as in the study group were applied. The accuracy of WE was compared between five commonly applied formulas using means of percentage errors (MPE), medians of absolute percentage errors (MAPE), and proportions of estimates within 10 % of actual birth weight.

RESULTS

In the whole study group, all equations showed a systematic underestimation of fetal weight (negative MPEs). Overall, best MAPE and MPE values were found with the Hadlock II formula, using BPD, AC and FL as biometric parameters (Hadlock II, MPE: -1.28; MAPE: 6.52). MPEs differed significantly between WE in the study and control group for all evaluated formulas: in the control group, either no systematic error (Hadlock III, IV and V) or a significant overestimation (Hadlock I, II) was found. Regarding MAPEs, application of the Hadlock III (HC, AC, FL) and V (AC) formula resulted in significant lower values in the control group (Hadlock III, MAPE: 7.48 vs. 5.95, p = 0.0008 and Hadlock V, MAPE: 8.79 vs. 7.52, p = 0.0085). No significant differences were found for the other equations.

CONCLUSIONS

A systematic underestimation of fetal weight has to be taken into account in sonographic WE performed intrapartum. Overall, the best results can be achieved with WE formulas using the BPD as the only head measurement.