Prediction of delivery date by sonography in the first and second trimesters

Human-guided deep learning with ante-hoc explainability by convolutional network from non-image data for pregnancy prognostication

BACKGROUND AND OBJECTIVE

Deep learning is applied in medicine mostly due to its state-of-the-art performance for diagnostic imaging. Supervisory authorities also require the model to be explainable, but most explain the model after development (post hoc) instead of incorporating explanation into the design (ante hoc). This study aimed to demonstrate a human-guided deep learning with ante-hoc explainability by convolutional network from non-image data to develop, validate, and deploy a prognostic prediction model for PROM and an estimator of time of delivery using a nationwide health insurance database.

METHODS

To guide modeling, we constructed and verified association diagrams respectively from literatures and electronic health records. Non-image data were transformed into meaningful images utilizing predictor-to-predictor similarities, harnessing the power of convolutional neural network mostly used for diagnostic imaging. The network architecture was also inferred from the similarities.

RESULTS

This resulted the best model for prelabor rupture of membranes (n=883, 376) with the area under curves 0.73 (95% CI 0.72 to 0.75) and 0.70 (95% CI 0.69 to 0.71) respectively by internal and external validations, and outperformed previous models found by systematic review. It was explainable by knowledge-based diagrams and model representation.

CONCLUSIONS

This allows prognostication with actionable insights for preventive medicine.

Lower accuracy in prediction of delivery date in Stockholm County following introduction of new guidelines

INTRODUCTION

In Stockholm County, new guidelines for predicting the day of delivery were introduced in 2010. Recent clinical use has indicated that the predictive quality might be suboptimal. This study compares the accuracy of three equations applied during the first (crown-rump length or bi-parietal diameter) and second trimesters (bi-parietal diameter or bi-parietal diameter combined with femur length).

MATERIAL AND METHODS

The accuracy of estimated delivery date was compared in 14 239 ultrasound exams using median deviations from actual birth date, proportion of postterms, births within ±seven days of estimated delivery date, accuracy variations depending on fetal size, alternative fetal measurements obtained on the same occasion, and menstrual age.

RESULTS

The bi-parietal diameter and crown-rump length formulae overestimated pregnancy length by two and three days respectively, causing 7-8% of pregnancies to be labelled postterm. A combined bi-parietal diameter+femur length formula overestimated by one day, with 5.1% postterms. No significant difference was found in the proportion of births within ±seven days. Second trimester estimated delivery date assessment had larger median variations than did first trimester assessment and suffered from shifting deviations across fetal size. The comparison of different biometry formulae in the same individual demonstrated one day extra deviation for bi-parietal diameter and three days extra deviation for crown-rump length compared with the combined bi-parietal diameter+femur length formula.

CONCLUSIONS

The algorithms and dating occasions tested seem inappropriate for the present 280-day term definition. Alternative formulae ought to be sought, and the assumed duration of pregnancy reconsidered; 283 days corresponds to the observed pregnancy length calculated from last menstrual period, and would better fit the observed results for first trimester ultrasound scans.

Sex-specific differences in fetal and infant growth patterns: a prospective population-based cohort study

Background

The objective of this study was to assess whether sex-specific differences in fetal and infant growth exist.

Methods

This study was embedded in the Generation R Study, a population-based prospective birth cohort. In total, 8556 live singleton births were included. Fetal growth was assessed by ultrasound. During the first trimester, crown-rump-length (CRL) was measured. In the second and third trimester of pregnancy head circumference (HC), abdominal circumference (AC) and femur length (FL) were assessed. Information on infant growth during the first 2 years of life was obtained from Community Health Centers and included HC, body weight and length.

Results

In the first trimester, male CRL was larger than female CRL (0.12 SD [95% CI 0.03,0.22]). From the second trimester onwards, HC and AC were larger in males than in females (0.30 SD [95% CI 0.26,0.34] and 0.09 SD [95% CI 0.05,0.014], respectively). However, FL in males was smaller compared to female fetuses (0.21 SD [95% CI 0.17,0.26]). Repeated measurement analyses showed a different prenatal as well as postnatal HC growth pattern between males and females. A different pattern in body weight was observed with a higher body weight in males until the age of 12 months where after females have a higher body weight.

Conclusions

Sex affects both fetal as well as infant growth. Besides body size, also body proportions differ between males and females with different growth patterns. This sexual dimorphism might arise from differences in fetal programming with sex specific health differences as a consequence in later life.

Electronic supplementary material

The online version of this article (doi:10.1186/s13293-016-0119-1) contains supplementary material, which is available to authorized users.

Women's perception of accuracy of ultrasound dating in late pregnancy: a challenge to prevention of prolonged pregnancy in a resource-poor Nigerian setting

Background

Expected date of delivery (EDD) is estimated from the last menstrual period (LMP) or ultrasound scan. Conflicts between these estimates especially on the part of the physician and his/her patient could pose a challenge to prevention of prolonged pregnancy. The objective of this study was to determine the perception and acceptability of menstrual dating (EDD derived from LMP) with regard to timing of labor induction for postdatism by pregnant women who have a late pregnancy (≥23 weeks’ gestation) ultrasound scan.

Methods

This cross-sectional study included 443 consecutive pregnant women receiving antenatal care at two tertiary health institutions in Enugu, Nigeria, from January 1, 2013 to March 31, 2013.

Results

The mean age of the women was 27.9±2.41 (range 17–45) years. Most ultrasound scans (90.8%, 357/389) were carried out in late pregnancy, and 41.9% (167/389) were self-referred. The majority of the respondents (51.7%, 229/443) did not accept induction of labor for postdatism at a certain menstrual dating-derived gestational age of 40 weeks plus 10 days if the late pregnancy ultrasound scan dating was less. Predictors of this poor attitude to timing of induction of labor for postdatism included low educational level, low social class, and poor knowledge of the limitations of ultrasound scan dating in late pregnancy (P<0.05).

Conclusion

The worrisome confidence in ultrasound scan dating is a challenge to the prevention of prolonged pregnancy and its complications in our environment. Antenatal health education should discourage self-referral for ultrasound scan dating and emphasize its limitations in late pregnancy as well as the perinatal effects of prolonged pregnancy.

Down syndrome maternal serum marker screening after 18 weeks of gestation: a countrywide study

OBJECTIVE

The objective of the study was to evaluate the efficacy of maternal serum markers in detecting Down syndrome after 18 weeks of gestation in women who book late for maternity care in a large national retrospective study.

STUDY DESIGN

During the period 2007-2012, 27,648 women, regardless of maternal age (17.4% were 35 years old and over), were included in a late Down syndrome screening program (18(+0) to 35(+6) weeks) using the maternal serum markers alpha-fetoprotein and human chorionic gonadotrophin-beta. Samples were assayed in a single laboratory. A dataset of median markers previously established in our laboratory was used for risk calculation. The control group consisted of 27,648 women (14(+0) to 17(+6) weeks) randomly selected from the routine database.

RESULTS

When the later screening group was compared with the standard second-trimester control group, the median multiples of medians (1.01 vs 0.98 for alpha-fetoprotein, 1.03 vs 0.98 for human chorionic gonadotrophin-beta), median risks (1 of 2414 vs 1 of 2720), false-positive rates (11.1% vs 11.6%), and trisomy 21 detection rates (83.3% vs 85.7%) did not differ significantly.

CONCLUSION

Late Down syndrome maternal serum screening is feasible with a good sensitivity/specificity compromise throughout gestation and is of clinical value in late-booking women.

Correction of systematic bias in ultrasound dating in studies of small-for-gestational-age birth: an example from the Iowa Health in Pregnancy Study

The authors examined whether early ultrasound dating (≤20 weeks) of gestational age (GA) in small-for-gestational-age (SGA) fetuses may underestimate gestational duration and therefore the incidence of SGA birth. Within a population-based case-control study (May 2002-June 2005) of Iowa SGA births and preterm deliveries identified from birth records (n = 2,709), the authors illustrate a novel methodological approach with which to assess and correct for systematic underestimation of GA by early ultrasound in women with suspected SGA fetuses. After restricting the analysis to subjects with first-trimester prenatal care, a nonmissing date of the last menstrual period (LMP), and early ultrasound (n = 1,135), SGA subjects' ultrasound GA was 5.5 days less than their LMP GA, on average. Multivariable linear regression was conducted to determine the extent to which ultrasound GA predicted LMP dating and to correct for systematic misclassification that results after applying standard guidelines to adjudicate differences in these measures. In the unadjusted model, SGA subjects required a correction of +1.5 weeks to the ultrasound estimate. With adjustment for maternal age, smoking, and first-trimester vaginal bleeding, standard guidelines for adjudicating differences in ultrasound and LMP dating underestimated SGA birth by 12.9% and overestimated preterm delivery by 8.7%. This methodological approach can be applied by researchers using different study populations in similar research contexts.

Preterm delivery risk in migrants in Italy: an observational prospective study

BACKGROUND

Various studies have ascertained different birth outcomes between resident and migrant populations in western countries. Considering preterm delivery (<37 complete weeks of gestation) as a perinatal risk condition, we assessed its rate in migrant and native Italian women who delivered in the main public hospital in Brescia (Italy).

METHODS

All migrant puerperas and a random sample of native puerperas hospitalized during the period February to May 2005 were included in the study after informed consent and filled in a self-administered multilanguage questionnaire enquiring about sociodemographic and obstetric data. Additional information including last menstrual period was obtained from personal obstetric records.

RESULTS

As many as 471 puerperas entered the study: 366 Italian and 105 migrant women coming from eastern Europe (41.9%), Asia (20%), South America (10.5%), and Africa (27.6%). Of the migrant population, 67 of 105 (63.8%) were at their first delivery in Italy (median interval from arrival: 3.8 y). Gestational age at delivery was assessed for 456 of 471 women (103 migrants and 353 Italians). A total of 36 (7.9%) preterm deliveries were registered: 22 (6.2%) in Italian and 14 (13.6%) in migrant puerperas (p value = 0.015). The highest preterm delivery rate was observed in African women (20.7%), while women from eastern Europe had a similar rate to Italians. In univariate analysis, factors associated to preterm delivery were parity and length of permanence in Italy. We could not demonstrate any correlation with smoking or with a delayed access to antenatal care (first obstetric evaluation after 12 complete weeks of gestation). In multivariate analysis, African origin was the only independent risk factor for preterm delivery [odds ratio (OR) = 3.54; p = 0.018].

CONCLUSIONS

In our setting, preterm delivery occurred more frequently in migrant women, particularly of African origin, and it is not associated to delayed access to antenatal care.