Advertisement

Prediction of labor outcome pilot study: evaluation of primiparous women at term

Open AccessPublished:August 12, 2022DOI:https://doi.org/10.1016/j.ajogmf.2022.100711

      Background

      Emergency operative delivery is associated with high fetal and maternal morbidity and mortality. It is of high importance to find means to predict the delivery mode before the onset of labor.

      OBJECTIVE

      This study aimed to investigate the potential of combined sonographic and clinical determination to predict the mode of delivery at term.

      STUDY DESIGN

      An observational prospective cohort study was deployed in a tertiary maternity hospital (Emergency County Hospital Craiova). Unselected low-risk primiparous pregnant women were evaluated weekly at term for ultrasound determinations (estimated fetal weight, head descent parameters, occiput posterior, cervical length), Bishop score, and maternal characteristics (age, height, weight). A thorough statistical analysis determined which variables were significantly correlated with the delivery mode.

      RESULTS

      Data from 276 term primiparous women were analyzed. Head descent parameters were strongly and significantly correlated with each other, but only progression distance was correlated with the delivery mode (gestational weeks 37, 38, 41, and the week before delivery). In the week before delivery, measurements of head-to-perineum distance and angle of progression reached almost significant P levels of.055 and.07, respectively. The following variables were significantly correlated with the delivery mode: body mass index in all term evaluations; progression distance for weeks 37 and 38; maternal age for week 39; Bishop score, estimated fetal weight, and occiput posterior for week 40; and body mass index, estimated fetal weight, and progression distance for the week before delivery. We also provided logistic regression equations for each week with correct delivery mode prediction, except for week 38. Cutoff values were established for each significant parameter per week. The cutoff values must be read in conjunction with the area under the curve, which ranged from 0.55 to 0.73, depending on the variable.

      CONCLUSION

      There are strong and significant correlations among the “head descent” ultrasound measurements at term. Body mass index is predictive of labor outcomes throughout term evaluations. Progression distance and body mass index measured at 37 to 38 weeks’ gestation correlate with the delivery mode and apparently can be used to forecast the delivery mode when the pregnancy reaches term.
      For the week before delivery, measurements of estimated fetal weight and progression distance can be used to forecast the delivery mode, perhaps as part of a policy for pregnant women with prelabor clinical signs.
      Larger studies with more data, particularly better-balanced data, are needed.

      Key words

      Why was this study conducted?

      This study aimed to build a statistical framework that predicts the labor outcome in primiparous women at term.

      Key findings

      The statistical framework showed that the following features are strongly and significantly correlated with the labor outcome: body mass index and the progression distance at 37 and 38 weeks of gestation; maternal age for week 39; Bishop score, estimated fetal weight, and occiput posterior for week 40; and body mass index, estimated fetal weight, and progression distance for the week before delivery.

      What does this add to what is known?

      Previous studies have aimed to predict the labor outcome for women from admission to the labor and delivery unit, that is, for women in active labor. Our study aimed to determine the features strongly and significantly correlated with labor outcome before the women enter the active labor phase.
      One of the goals of modern obstetrics is the prediction of the delivery mode, preferably before the onset of labor, because emergency operative delivery is associated with high fetal and maternal morbidity and mortality, and patient perception of personal failure and/or suboptimal care.
      • Dietz HP
      • Lanzarone V
      • Simpson JM.
      Predicting operative delivery.
      Besides clinical examinations, many studies used ultrasound (US) measurements at term, before labor onset to predict the delivery mode.
      • Levy R
      • Zaks S
      • Ben-Arie A
      • Perlman S
      • Hagay Z
      • Vaisbuch E.
      Can angle of progression in pregnant women before onset of labor predict mode of delivery?.
      • Gillor M
      • Vaisbuch E
      • Zaks S
      • Barak O
      • Hagay Z
      • Levy R.
      Transperineal sonographic assessment of angle of progression as a predictor of successful vaginal delivery following induction of labor.
      • Dietz HP
      • Moore KH
      • Steensma AB.
      Antenatal pelvic organ mobility is associated with delivery mode.
      • Rane SM
      • Guirgis RR
      • Higgins B
      • Nicolaides KH.
      Pre-induction sonographic measurement of cervical length in prolonged pregnancy: the effect of parity in the prediction of the need for Cesarean section.
      • Roman H
      • Verspyck E
      • Vercoustre L
      • et al.
      Does ultrasound examination when the cervix is unfavorable improve the prediction of failed labor induction?.
      • Rane SM
      • Guirgis RR
      • Higgins B
      • Nicolaides KH.
      The value of ultrasound in the prediction of successful induction of labor.
      • Rane SM
      • Guirgis RR
      • Higgins B
      • Nicolaides KH.
      Models for the prediction of successful induction of labor based on pre-induction sonographic measurement of cervical length.
      • Dietz HP
      • Lanzarone V.
      Measuring engagement of the fetal head: validity and reproducibility of a new ultrasound technique.
      • Eggebø TM
      • Gjessing LK
      • Heien C
      • et al.
      Prediction of labor and delivery by transperineal ultrasound in pregnancies with prelabor rupture of membranes at term.
      • Dupuis O
      • Silveira R
      • Zentner A
      • et al.
      Birth simulator: reliability of transvaginal assessment of fetal head station as defined by the American College of Obstetricians and Gynecologists classification.
      • Oláh KS.
      Reversal of the decision for caesarean section in the second stage of labour on the basis of consultant vaginal assessment.
      • Buchmann E
      • Libhaber E.
      Interobserver agreement in intrapartum estimation of fetal head station.
      • Wolpert DH
      • Macready WG.
      No free lunch theorems for optimization.
      A variety of sonographic parameters were previously proposed in the literature, alone or in combination with clinical features, regarding fetal head engagement in the maternal pelvis (head progression distance [PD], head-to-perineum distance [HPD], angle of progression [AOP]), occiput posterior, estimated fetal weight (EFW), cervical length and angle, and pelvic organ mobility. Nevertheless, only a few studies attempted to estimate the labor outcome before uterine contraction onset or rupture of membranes, and encouraging data were reported. Dietz et al
      • Eggebø TM
      • Gjessing LK
      • Heien C
      • et al.
      Prediction of labor and delivery by transperineal ultrasound in pregnancies with prelabor rupture of membranes at term.
      found that a model that combines clinical and US variables is likely to predict delivery mode accurately in 87% of nulliparous women, whereas Levy et al found that a narrow angle of progression (<95°) in nonlaboring nulliparous women at term is associated with a high rate of cesarean delivery. An important issue is that we lack confirmation for these findings, and this represents an evidence gap that our study aimed to fill. Conversely, we aimed to extend the studied variables for better predictions and to investigate their value throughout term to identify the best moment for evaluation.
      Our research consisted of longitudinal assessment of fetomaternal characteristics at term using weekly US scans, including the fetal head descent and occiput posterior, EFW, and cervical length. Concomitantly, clinical parameters were recorded, namely maternal characteristics and the Bishop score.
      Several original approaches were used in our study. First, we aimed for a combined US and clinical analysis that takes into account all the parameters that were considered predictive in previous research. Second, the sequential design of the study enabled the analysis of predictions in early or full-term pregnancy. Third, we could determine the added value of each US and clinical feature at any gestational age at term, including the week before delivery measurements.
      The aim of this observational study was to determine which parameters measured in nulliparous women at term before labor onset may serve for the prediction of labor outcome using logistic regression. A thorough statistical analysis was performed for this purpose.

      Methods and analysis

      Setting, participants, recruitment

      This observational prospective cohort study was deployed in a tertiary maternity hospital (Emergency County Hospital Craiova). Data analysis was carried out by the Department of Computer Science, Faculty of Sciences, University of Craiova.
      Primiparous pregnant women admitted to the Prenatal Unit for the third-trimester well-being scan at ≥37 gestational weeks were considered eligible for the study. Pregnancies with indications for elective cesarean delivery, noncephalic presentation, multiple pregnancies, previous cesarean delivery, fetal growth restriction, preeclampsia, and diabetes mellitus were excluded from the study. The eligible women were invited to take part in the study for a series of weekly scans and clinical examinations at term. Gestational age was determined according to early pregnancy US dating. The patients were recruited consecutively, depending on the availability of a sonographer with experience in transperineal US (TPU).

      Procedures

      During the standard consultations in the Prenatal Diagnostic Unit, the sonographer provided brief information about the research to eligible cases and invited the patient to take part in the study. Written informed consent was obtained if the patient agreed to study participation and fulfilled the inclusion criteria.
      US and clinical evaluations were started at the first presentation at term.
      The sonographic examinations were performed by obstetricians with appropriate training in transabdominal and transperineal obstetrical US and a minimum of 2 years of experience. Image acquisition was conducted using GE Voluson 730 Pro (GE Medical Systems, Zipf, Austria) and LOGIQ e (GE Healthcare, Shanghai, China) US machines, equipped with 2–5-MHz, 4–8-MHz, and 5–9-MHz curvilinear transducers.
      The following sonographic planes were obtained for fetomaternal measurements:
      • 1.
        biparietal diameter and head circumference (HC), abdominal circumference, femur diaphysis length planes (for fetal biometry and fetal weight estimation)
      • 2.
        transabdominal suprapubic transverse plane (to determine occiput posterior)
      • 3.
        infrapubic or translabial sagittal plane in the semirecumbent posterior, with legs flexed (for fetal head descent evaluation)
      • 4.
        transperineal transverse plane, at the level of the ischial tuberosity, applying firm pressure without creating discomfort, with the transducer moved and angled until the shortest distance to the fetal skull was visualized (used to evaluate the fetal HPD)
      • 5.
        transvaginal sagittal view of the cervix, avoiding cervical distortion (for cervical length measurement)
      The measurements were performed using the techniques described in the literature:
      • EFW was calculated on the basis of the fetal biometry measurements.
      • Occiput posterior was determined transabdominally and recorded on a data sheet depicting a circle, like a clock, divided into 24 sections, each of 15°. The position of the occiput was classified accordingly as occiput anterior (OA), occiput posterior (OP), left or right occiput transverse, left or right OA, or left or right OP.
      • Fetal head descent parameters including AOP, PD, direction angle (DA), and fetal HPD were measured as described previously in the literature
        • Barbera AF
        • Pombar X
        • Perugino G
        • Lezotte DC
        • Hobbins JC.
        A new method to assess fetal head descent in labor with transperineal ultrasound.
        • Kalache KD
        • Dückelmann AM
        • Michaelis SA
        • Lange J
        • Cichon G
        • Dudenhausen JW.
        Transperineal ultrasound imaging in prolonged second stage of labor with occipitoanterior presenting fetuses: how well does the ‘angle of progression’ predict the mode of delivery?.
        • Molina FS
        • Terra R
        • Carrillo MP
        • Puertas A
        • Nicolaides KH.
        What is the most reliable ultrasound parameter for assessment of fetal head descent?.
        • Ramanathan G
        • Yu C
        • Osei E
        • Nicolaides KH.
        Ultrasound examination at 37 weeks’ gestation in the prediction of pregnancy outcome: the value of cervical assessment.
        (Figure 1).
        Figure 1
        Figure 1Ultrasound procedures and example images for the measurement of head descent parameters
        Iliescu. Statistical framework for labor outcome prediction. Am J Obstet Gynecol MFM 2022.
        A, Placement of the ultrasound probe in the infrapubic translabial sagittal plane. B, Measurement of the progression angle between the long axis of the pubic symphysis and a line extending from its most inferior portion tangentially to the fetal skull (orange lines); measurement of the direction angle according to Iliescu et al33 (with green, biparietal line perpendicular to the major longitudinal axis of the fetal head); measurement of the progression distance as the minimal distance between the infrapubic line and the leading part of the fetal skull (red arrow). C, Placement of the ultrasound probe in the infrapubic transverse plane. D, Measurement of the head-to-perineum distance as the shortest distance from the skin surface of the perineum to the outer bony limit of the fetal skull.
      • Cervical length was measured as the distance between the external and internal os, the furthest points at which the cervical walls were juxtaposed.
        • Tutschek B
        • Braun T
        • Chantraine F
        • Henrich W.
        A study of progress of labour using intrapartum translabial ultrasound, assessing head station, direction, and angle of descent.
      Clinical evaluations were performed immediately after each US scan. The pregnant women were clinically assessed by a senior consultant blinded to the US measurements that noted the Bishop score.
      The clinical and US evaluations were performed weekly until birth. There were 16 cases lost to follow-up that were excluded from the study.
      The following maternal characteristics were noted: age, height, weight, and gestational age. Labor characteristics included: mode and date of delivery, fetal Apgar score and birthweight, whether labor was spontaneous or induced, and use of oxytocin or epidural anesthesia.
      Given the difficulties of obtaining correct transperineal measurements on a nonengaged head, all sonographers completed a 1-day workshop and participated in group supervision sessions in the first month of the study to ensure protocol fidelity. The midline structures are easily apparent at lower stations during labor when the transducer is oriented in a standard fashion, parallel to the pubic symphysis, because the structures are not shadowed by the pubic bone. However, in our setting, before active labor and at higher head stations, the midline is safely visualized if the transducer is oriented obliquely and caudal to the symphysis long axis in the same plane.
      • Dietz HP
      • Lanzarone V.
      Measuring engagement of the fetal head: validity and reproducibility of a new ultrasound technique.
      ,
      • Eggebø TM
      • Gjessing LK
      • Heien C
      • et al.
      Prediction of labor and delivery by transperineal ultrasound in pregnancies with prelabor rupture of membranes at term.
      A different insonation angle, but in the same sagittal perineal infrapubic plane, cannot alter the results of intrapartum TPU measurements because the relation between the pubic bone and the fetal head remains the same.
      Posttraining, the first examinations were supervised to ensure that the technique and measurements are appropriate. Periodic quality checks were performed during which all the evaluations were reviewed by 1 experienced sonographer with a special interest in TPU (D.I.).
      Both the patient and the labor and delivery team were blinded to the weekly measurements to prevent being influenced by knowledge of previous findings.
      Our primary outcome was to investigate the potential of combined sonographic and clinical determinations to predict the mode of delivery at term on the basis of data acquired during weekly evaluations.

      Statistical analysis

      In this study, we performed a thorough statistical analysis to determine which variables are indeed strong and significantly correlated with the output. Before performing any test, we first computed the required sample size using power analysis. Hence, by computing the appropriate sample size given a certain statistical power, we could reliably trust the obtained results. Grosso modo, if our tests revealed significant differences between the 2 groups (vaginal vs cesarean delivery) in a small given number of subjects, we could be sure with a given probability (the power of the test) that the same difference would be found in a larger cohort. For a statistical power >95%, with type I error α=0.05, we computed a sample size of 246. Taking into account that approximately 20% of deliveries are cesarean deliveries, for a statistical power >95% we would need 196 vaginal and 50 cesarean delivery cases. Therefore, the results obtained in our dataset that contained 276 samples with 221 vaginal and 55 cesarean deliveries achieved a very good statistical power. A drawback of the dataset was its lack of balance, hence the model tended to be sensitive to false-positives.
      The statistical analysis was 3-fold. At first, we were interested in determining which variables were correlated with the delivery mode per week. Hence, for each week we computed the correlation matrix and applied the logistic regression to determine the predictive variables. After discovering which variables were important, we wanted to determine whether there were significant differences between the delivery modes in terms of these variables. For this, we applied the t test for independent variables. We also provided concrete examples of how to use the logistic regression equations for each week. We checked whether there were statistically significant differences between the same variables throughout the weeks using the Friedman analysis of variance (ANOVA) and Kendall's concordance coefficient, which revealed a chi-square value of 10.77 with a P level of.004.
      The final step in our statistical analysis was to find a reliable and valid cutoff point for classifying cases for each week and variable. We used several statistical methods: receiver operating characteristic (ROC) curves with the G-mean and Youden's J statistic, precision-recall curves, and the F1-score to establish the threshold for each significant variable per week. To produce even more reliable results, we used threshold tuning.

      Results

      Data from 276 term primiparous women that delivered in our unit during 2019 were analyzed (Figure 2). Participant recruitment and investigation are presented in Figure 1. Maternal and labor characteristics are presented in Table 1.
      Figure 2
      Figure 2Flow chart summarizing the study group
      Iliescu. Statistical framework for labor outcome prediction. Am J Obstet Gynecol MFM 2022.
      AOP, angle of progression; EFW, estimated fetal weight, GW, gestational weeks; HPD, head-to-perineum distance; PD, progression distance.
      Table 1Characteristics of study population of 276 nulliparous women at term and labor outcome
      Iliescu. Statistical framework for labor outcome prediction. Am J Obstet Gynecol MFM 2022.
      Maternal and labor characteristicsInterval, mean/median
      Maternal age18–35, 27.74
      Maternal weight52–109, 72.56
      Maternal height150–189, 167.21
      Maternal BMI20.07–38.37, 25.97
      Maternal ethnicity276 White
      Gestational age at inclusion37 wk+0 d–41 wk+2 d
      Gestational age at delivery37 wk+1 d–42 wk+1 d, 39.5
      Mode of delivery220 vaginal deliveries (79.7%)

      56 cesarean deliveries (20.3%)
      Apgar score5–10, 9
      Birthweight2530–3960, 3395, 3400 g
      Spontaneous and induced labor256 spontaneous (92.8%), 20 induced (7.2%)
      Spontaneous vaginal delivery or instrumental29 instrumental (13.2% of vaginal births)
      BMI, body mass index.

      Predictive variables for delivery mode

      For each week, we computed the correlation matrix and applied the logistic regression to determine the predictive variables. Table 2 depicts the predictor variables that surpassed the significance level (P>.05), whereas Figure 3, A to F presents the correlation matrix heatmaps for each week, and for the week before delivery.
      Table 2Statistically significant predictor variables per week and equality of variances for weeks 37 to 41 and the week before delivery
      Iliescu. Statistical framework for labor outcome prediction. Am J Obstet Gynecol MFM 2022.
      WeekVariableP value for statistically significant predictor variablesF-value
      37BMI.4650.535
      PD.1112.535
      38BMI.9650.0019
      PD.9620.0022
      39Maternal age.2161.540
      BMI.9650.0019
      40EFW.8780.0236
      BMI.2961.107
      OP.3320.329
      41BMI.5160.421
      EFW.3720.798
      PD.640.206
      Before deliveryBMI.5160.421
      EFW.3720.798
      PD.640.206
      BMI, body mass index; EFW, estimated fetal weight; OP, occiput posterior; PD, progression distance.
      Figure 3
      Figure 3Correlation matrix heatmaps
      Iliescu. Statistical framework for labor outcome prediction. Am J Obstet Gynecol MFM 2022.
      For weeks A, 37, B, 38, C, 39, D, 40, E, 41, and F, week before delivery.
      Figure 3, A to F and Table 2 show that the parameters AOP, DA, PD, and HPD were strongly and significantly correlated with each other, but only PD was correlated with the output/type of birth (weeks 37, 38, 41, and the week before delivery), and DA in week 41. This is not entirely surprising given that AOP, DA, PD, and HPD are linearly dependent on each other, whereas PD and DA are nonlinearly dependent on the type of birth.
      The P level measuring the significance of correlation of AOP, DA, PD, and HPD with the type of birth is presented in Table 3. In the week before delivery, PD was significantly correlated with the type of birth, but HPD and AOP parameters had almost significant P levels of.055 and.07, respectively. In our case, with an appropriate sample size our statistical power was high (95%), hence we could use the classical threshold of.05 for the P level.
      • Debby A
      • Rotmensch S
      • Girtler O
      • Sadan O
      • Golan A
      • Glezerman M.
      Clinical significance of the floating fetal head in nulliparous women in labor.
      • Shin KS
      • Brubaker KL
      • Ackerson LM.
      Risk of cesarean delivery in nulliparous women at greater than 41 weeks’ gestational age with an unengaged vertex.
      • Oboro VO
      • Tabowei TO
      • Bosah JO.
      Fetal station at the time of labour arrest and risk of caesarean delivery.
      • Altman DG.
      Practical statistics for medical research.
      Table 3The significance of correlation (P level) of head descent ultrasound parameters with the type of birth
      Iliescu. Statistical framework for labor outcome prediction. Am J Obstet Gynecol MFM 2022.
      WeekParameters/P level
      AOPDAPDHPD
      37.876.239.038.156
      38.824.6.037.116
      39.242.245.431.407
      40.566.59.9.44
      41.642.004.049.938
      Before delivery.07.844.0004.055
      AOP, angle of progression; DA, direction angle; HPD, head-to-perineum distance; PD, progression distance.
      Besides the significance, we also depicted the strong correlations between the variables and the delivery mode through the correlation matrix heatmaps (Figure 3).

      Significance of the predictive variables

      We applied the t test for independent variables after verifying 2 assumptions: normal distribution and equal variances.
      In terms of normality, with a sample size of >30 for week 37, because of the central limit theorem we could assume that the sample's distribution was approximately normal. For the equality of variances, we applied Levene's test. The results are reported in Table 2 (F-values).
      For the weeks 37 to 41 and for the week before delivery, the samples had equal variances. For weeks 37, 38, and 39, we did not have to verify the samples’ distributions because their sizes were >30. Therefore, we could proceed with applying a t-test for the significant variables. The results are presented in Table 4, from which we can draw the following conclusions.
      Table 4The T test for independent variables for weeks 37 to 41 and for the week before delivery in comparison with independent groups (t test and Mann–Whitney U test) for week 40
      Iliescu. Statistical framework for labor outcome prediction. Am J Obstet Gynecol MFM 2022.
      WeekVariablet-valueU/ZP value
      37BMI1.801.0733
      PDa1.995a.047a
      38BMIa2.229a.027a
      PD0.481.630
      39Maternal age−1.513.132
      BMIa2.229a.027a
      40EFWa−2565a.012a
      BMI−1.311.194
      OP2.0/−0.288.722
      Week before deliveryBMIa2.876a.004a
      EFW−1.077.282
      PDa2.423a.016a
      BMI, body mass index; EFW, estimated fetal weight; OP, occiput posterior; PD, progression distance, U/Z, U statistics/ Z statistics.
      a statistically significant.
      For week 37 measurements:
      • PD was strongly correlated with the delivery mode.
      • Body mass index (BMI) and PD were significantly correlated with the delivery mode.
      • The t-test revealed that there were no significant differences between the 2 independent groups (vaginal vs cesarean deliveries) in terms of the BMI variable, that is, this predictor could not be used to determine the delivery mode in week 37, whereas the PD variable showed that there were significant differences between the 2 delivery modes in week 37. Thus, the PD measurement could be used to forecast the delivery mode for week 37.
      • The logistic regression equation for week 37 was:
        logit(p)=0.6890.18×BMI0.55×PD


      For week 38 measurements:
      • BMI was strongly correlated with the delivery mode.
      • BMI and PD were significantly correlated with the delivery mode.
      • Contrary to results for week 37, the t-test revealed that there were no significant differences between the 2 independent groups (vaginal vs cesarean deliveries) in terms of the PD variable, that is, this predictor could not be used to determine the delivery mode in week 38, whereas the BMI variable showed that there were significant differences between the 2 delivery modes in week 38. Thus, the PD measurement could be used to forecast the delivery mode for week 37 but not for week 38.
      • The logistic regression equation for week 38 was:
        logit(p)=0.6370.21×BMI0.55×PD


      For week 39 measurements:
      • Maternal age was strongly correlated with the delivery mode.
      • Maternal age and BMI were significantly correlated with the delivery mode.
      • Again, as in the case of week 38, the t-test revealed that there were significant differences in terms of mean BMI between the 2 groups.
      • The logistic regression equation for week 39 was:
        logit(p)=0.797+0.223×Age0.25×BMI


      Because we did not have a large sample of subjects that underwent cesarean delivery in week 40, we performed a normality test to determine whether the distribution was nearly Gaussian. There was no need for this in the vaginal delivery group, with the sample size being >30. Thus, we applied the Kolmogorov–Smirnov (K-S) and the Lilliefors test with the Shapiro–Wilk W (S-W) test for the cesarean delivery group. Table 6 depicts the normality test results.
      Table 5The t test for independent variables for weeks 37 to 41 and for the week before delivery
      Iliescu. Statistical framework for labor outcome prediction. Am J Obstet Gynecol MFM 2022.
      WeekVariablet-valueP value
      37BMI1.801.0733
      PDa1.995a.047a
      38BMIa2.229a.027a
      PD0.481.630
      39Maternal age−1.513.132
      BMIa2.229a.027a
      BMI, body mass index; PD, progression distance.
      a statistically significant.
      Table 6Normality tests for week 40 (cesarean delivery group) and week 41
      Iliescu. Statistical framework for labor outcome prediction. Am J Obstet Gynecol MFM 2022.
      Week 40 (cesarean delivery group)
      VariableK-SLilliefors PS-WP
      EFW0.135>0.200.959.718
      BMI0.126>0.200.958.051
      OP
      statistically significant.
      0.267
      statistically significant.
      <0.01
      statistically significant.
      0.845
      statistically significant.
      .019
      statistically significant.
      Week 41
      statistically significant.
      VariableK-SLilliefors PS-WP
      BMI (vaginal)
      statistically significant.
      0.206
      statistically significant.
      <0.01
      statistically significant.
      0.868
      statistically significant.
      .0048
      statistically significant.
      EFW (vaginal)0.091>0.20.975.803
      PD (vaginal)0.144>0.20.959.427
      DA (vaginal)0.123>0.20.971.7
      BMI (cesarean delivery)0.176>0.20.999.981
      EFW (cesarean delivery)
      statistically significant.
      0.384
      statistically significant.
      <0.05
      statistically significant.
      0.75
      statistically significant.
      .0000
      statistically significant.
      PD (cesarean delivery)
      statistically significant.
      0.367
      statistically significant.
      <0.1
      statistically significant.
      0.791
      statistically significant.
      .094
      statistically significant.
      DA (cesarean delivery)
      statistically significant.
      0.371
      statistically significant.
      <0.1
      statistically significant.
      0.782
      statistically significant.
      .073
      statistically significant.
      BMI, body mass index; DA, direction angle; EFW, estimated fetal weight; K-S, Kolmogorov–Smirnov test; OP, occiput posterior; PD, progression distance; S-W, Shapiro–Wilk W test.
      a statistically significant.
      As shown, the EFW and BMI samples were normally distributed for both the vaginal and the cesarean delivery group, whereas the OP sample for cesarean delivery was not normally distributed; thus, we could not use the t test for the independent group. The nonparametric alternative is represented by the Mann–Whitney U test, which we applied. Because the equality of variances was already checked, we could proceed and apply these tests. The results are presented in Table 4.
      For the variables measured during week 40, the statistical analysis revealed:
      • Bishop score, EFW, and OP were strongly correlated with the delivery mode.
      • EFW, BMI, and OP were significantly correlated with the delivery mode.
      • The t-test revealed that there were significant differences in mean EFW and no differences in BMI, whereas the Mann–Whitney U test showed no significant difference in OP.
      • The logistic regression equation for week 40 was:
        logit(p)=2.2+0.311×EFW0.11×BMI0.35×OP


      Regarding the statistical analysis for week 41, in this week the sample size was insufficient for either delivery mode. Thus, we had to apply normality tests for all 4 significant variables (BMI, EFW, PD, and DA). Table 6 presents the results for both the K-S and S-W tests.
      The normality tests revealed that the BMI variable in the vaginal group and the EFW, PD, and DA variables in the cesarean delivery group were not normally distributed. The t test could not be applied in these conditions, thus we applied the Mann–Whitney U test. For all the comparisons, the U statistics equaled 0, with a P level of 1.000. This indicated that all values from the first group of samples were greater than those from the second group, with no statistical significance.
      The logistic regression equation for week 41 was:
      logit(p)=1.3130.38×BMI+0.42×GE+0.89×PD0.61×DA


      Finally, regarding the measurements performed during the week before delivery, we did not have to verify the samples’ distributions because both of their sizes were >30. Having both assumptions regarding the samples’ distribution and the equality of variances fulfilled, we could apply the t test for independent samples. The results are presented in Table 4.
      For the week before delivery evaluations, we could draw the following conclusions (Table 5):
      • None of the variables were strongly correlated with the delivery mode.
      • Nevertheless, BMI, EFW, and PD were significantly correlated with the delivery mode.
      • The t-test revealed that there were no significant differences between the 2 independent groups (vaginal vs cesarean deliveries) in terms of the EFW variable, that is, this predictor could not be used to determine the delivery mode in the week before delivery. Here we reached a contradiction between 2 procedures that determined the significance of a variable. EFW was significant according to the t-test but not significant according to logistic regression. This indicated that the relationship was sufficiently nonlinear, thus only the extension of logistic regression to nonlinear relationships could detect and quantify this situation, which the t test could not.
      • The BMI and PD variables showed that there were significant differences between the 2 delivery modes in the week before delivery.
      • The logistic regression equation for the week before delivery was:
        logit(p)=0.1030.3×BMI+0.141×EFW0.57×PD


      In Table 7, we also provided concrete examples of how to use the logistic regression equations for each week. To determine the result we transformed the value of the logit(p) into the probability p using the following formula:
      p=11+elogit(p).


      With p as the probability of a patient giving birth vaginally, the patient would most probably give birth through cesarean delivery if P<.5; otherwise, she would give birth vaginally.
      Table 7Logistic regression examples per week
      Iliescu. Statistical framework for labor outcome prediction. Am J Obstet Gynecol MFM 2022.
      WeekVar 1Var 2Var 3Var 4Regression equation per weekResultGround truth
      37BMIPDlogit(p)=0.6890.18×23.050.55×0.41=6.860.00100
      38BMIPDlogit(p)=0.6370.21×23.050.55×0.41=4.420.0101
      39AgeBMIlogit(p)=0.797+0.223×260.25×24.800.5911
      40EFWBMIOPlogit(p)=2.2+0.311×3.20.11×25.0.35×1=4.300.0100
      41BMIEFWPDDAlogit(p)=1.3130.38×24.91+0.42×29350.89×0.20.61×57.69=1186.550.2600
      Before deliveryBMIEFWPDlogit(p)=0.1030.3×21.63+0.141×3024+0.57×2.23=421.270.2600
      BMI, body mass index; DA, direction angle; EFW, estimated fetal weight; OP, occiput posterior; PD, progression distance; Var, variable.
      As shown in Table 6, from the 5 examples provided for each week, 4 correctly forecasted the delivery mode as vaginal or cesarean, proving that the chosen variables were indeed strong and significantly correlated with the delivery mode. The only false prediction was found in the example for week 38.

      Statistically significant differences throughout the weeks

      Because the most significant variables were PD and BMI, we explored how they changed through weeks 37 to 39. Hence, we applied the Friedman ANOVA and Kendall's concordance coefficient, which revealed for PD a chi-square value of 10.77 with a P level of.004, and for BMI a chi-square value of 9.84 with a P of.071. We can draw the following conclusion: the PD variable changed through the weeks, thus even if it was significant in the first 2 weeks (37 and 38), its significance was lost in week 39, whereas BMI did not change significantly through the weeks. We could not apply the test for weeks 40 and 41 because of the insufficient number of samples.

      Cutoff analysis

      We used several statistical methods to find a reliable and valid cutoff point for classifying cases for each week and variable: ROC curve, G-mean, Youden's J statistic, the precision-recall curve, and F1-score. The G-mean is an unbiased evaluation metric for imbalanced classification, being the geometric mean of sensitivity (recall) and specificity. We also used Youden's J statistic to determine the best threshold for the classification. Finally, we implemented a looping mechanism to find the optimal threshold for maximizing the F1-score as an unbiased metric.
      In Figure 4, we plotted the data distribution for weeks 37 to 40 for the significant variables. Week 41 had a very unbalanced dataset, which would not have provided any reliable insights. BMI distribution changed through the weeks as new women entered the study or as others gave birth. Besides the data distribution, we added a box-and-whisker plot for each significant variable per week.
      Figure 4
      Figure 4Data distribution and box-and-whiskers plots for significant variables for weeks 37 to 40
      Iliescu. Statistical framework for labor outcome prediction. Am J Obstet Gynecol MFM 2022.
      A, PD for week 37, B, BMI for week 37, C, PD for week 38, D, BMI for week 38, E, age for week 39, F, BMI for week 39, G, EFW for week 40, H, OP for week 40, and I, BMI for week 40.
      BMI, body mass index; EFW, estimated fetal weight, OP, occiput posterior; PD, progression distance.
      We used the ROC curves with the G-mean and Youden's J statistics, precision-recall curves, and F1-score to establish the threshold for each significant variable per week (Figure 4). To produce even more reliable results, we used threshold tuning (Figure 5). The obtained results with the cutoff values are presented in Table 7.
      Figure 5
      Figure 5ROC curve and G-mean optimal threshold for significant variables for weeks 37 to 40
      Iliescu. Statistical framework for labor outcome prediction. Am J Obstet Gynecol MFM 2022.
      A, PD for week 37, B, BMI for week 37, C, PD for week 38, D, BMI for week 38, E, age for week 39, F, BMI for week 39, G, EFW for week 40, H, OP for week 40, and I, BMI for week 40.
      BMI, body mass index; EFW, estimated fetal weight, OP, occiput posterior; PD, progression distance; ROC, receiver operating characteristic.

      Discussion

      The main finding of our study was that some clinical and US term measurements were correlated with labor outcome. From the variables measured at 37 and 38 weeks, PD and BMI were significantly and strongly correlated with the delivery mode. Thus, these measurements apparently can be used to forecast the delivery mode when the pregnancy reaches term.
      BMI remained predictive in our group beyond 38 weeks of gestation, as did the women's age (at 39 weeks’ gestation) and EFW and OP (both at 40 weeks). These maternal and fetal characteristics are known to influence labor outcome, but it is hard to explain their inconstant significance during our longitudinal prelabor assessment. Obviously, a larger sample is needed. However, our preliminary results showed the potential of these fetal and maternal parameters to estimate the labor outcome.
      BMI, EFW, and PD were significantly correlated with the delivery mode in the week before delivery, and the t-test indicated that EFW and PD measurements can be used to forecast the delivery mode. The only reasonable question is how can we determine that week? The answer is that we cannot the determine which week will be the week before delivery, and the only way to have less than a week data, based on the weekly examinations at term. This would not be realistic as standard practice, but it could apply in cases with prelabor clinical signs or in selected pregnancies at the couple's request.
      There were strong and significant correlations between the “head descent” US measurements (AOP, DA, PD, and HPD). Their correlation was previously demonstrated during labor,
      • Shin KS
      • Brubaker KL
      • Ackerson LM.
      Risk of cesarean delivery in nulliparous women at greater than 41 weeks’ gestational age with an unengaged vertex.
      but this was not extensively investigated before labor onset. Only PD was correlated with the delivery mode (measurements from weeks 37, 38, 41, and the week before delivery). An explanation for the absence of strong correlations with the other US head descent measurements would be the fact that AOP, DA, PD, and HPD were linearly dependent on each other, whereas PD was nonlinearly dependent on the type of birth.
      Furthermore, we attempted to understand why PD was not constantly predictive regarding the labor outcome throughout the term period. The study of the PD dataset showed us that PD increased with gestational age at term, and during weeks 39 and 40 the statistical significance was not reached anymore. In other words, fetal head positioning seems to be completed by this gestational age, and other fetal–maternal variables remain (BMI) or become important (maternal age, Bishop schore, fetal weight, and occiput posterior). Nevertheless, fetal positioning is important at full term (41 gestational weeks), when fetal head engagement measured by PD becomes relevant. In addition, PD is predictive when measured the week before delivery, which confirms the proposition that a good positioning of the fetal head before labor is important for labor outcome, whereas fetal head nonengagement is associated with an unfavorable prognosis for vaginal delivery.
      • Shin KS
      • Brubaker KL
      • Ackerson LM.
      Risk of cesarean delivery in nulliparous women at greater than 41 weeks’ gestational age with an unengaged vertex.
      • Oboro VO
      • Tabowei TO
      • Bosah JO.
      Fetal station at the time of labour arrest and risk of caesarean delivery.
      • Altman DG.
      Practical statistics for medical research.
      Moreover, Lipshuetz et al found that sonographic fetal HC ≥35 cm, measured within the week before delivery, is an independent risk factor for unplanned cesarean delivery, whereas both fetal HC ≥35 cm and EFW ≥3900 g significantly increased the risk of a prolonged second stage of labor.
      • Hodge JK
      • Klima RE.
      The mathematics of voting and elections: a hands-on approach, Mathematical world series 22.
      We noted that in the week before delivery, PD was the only US parameter significantly correlated with the type of birth. We should underline the fact that here we encountered an interesting result regarding the cutoff value of the P level, which is arbitrary and offers just an assumption. The HPD and AOP parameters had P=.055 and P=.07, respectively. Was this a good enough reason to reject them in our analysis? The difference between the 2 values was rather small. How could we ensure that we do not misinterpret the data? In our case, with an appropriate sample size our statistical power was high (95%), hence we could use the classical threshold of.05 for the P level.

      Tannenbaum P. Excursions in modern mathematics. 6th ed. Upper Saddle River, NJ: Prentice Hall.

      • Belciug S.
      Artificial Intelligence in Cancer-diagnostic to tailored treatment.
      • Rizzo G
      • Aiello E
      • Bosi C
      • D'Antonio F
      • Arduini D
      Fetal head circumference and subpubic angle are independent risk factors for unplanned cesarean and operative delivery.
      • Ghi T
      • Eggebø T
      • Lees C
      • et al.
      ISUOG Practice Guidelines: intrapartum ultrasound.
      A previous study
      • Levy R
      • Zaks S
      • Ben-Arie A
      • Perlman S
      • Hagay Z
      • Vaisbuch E.
      Can angle of progression in pregnant women before onset of labor predict mode of delivery?.
      found that a narrow AOP (<95°) in nulliparous women who delivered within 1 week of sonography was associated with a high rate of cesarean delivery. We could not confirm this finding, but this parameter almost reached significance in our “week before delivery” group. However, we should underline the population and study design differences between the 2 studies: Levy et al included women presenting at the maternal–fetal medicine unit at 39 to 42 completed weeks of gestation with contractions, but not yet in active labor, whereas we included unselected women with term pregnancies, monitored weekly, and not admitted for contractions or prelabor.
      The obtained cutoff values must be read in conjunction with the area under the curve obtained for each variable, which ranged from 0.55 to 0.73 (Figure 6). A cutoff value of 0.55 was considered a failure, providing a 50–50 chance of predicting a vaginal vs cesarean delivery, whereas a cutoff value of 0.73 provided a 73% chance of distinguishing between vaginal and cesarean delivery (Table 8). Given the lack of exact prediction, it would be difficult to recommend a primary cesarean delivery on the basis of any of the parameters, even those that reached clinical significance. This is why we attempted to integrate the features found to be independently associated with vaginal or cesarean delivery by building a multiparametric prediction models. This is not new for maternal–fetal medicine, in a similar way the first-trimester combined test is designed to determine the pregnancy genetic risk. Thus, we provided logistic regression equations that correctly forecast the delivery mode as vaginal or cesarean, except in week 38. Because the logistic regression is a stochastic algorithm, not a deterministic one, its outcome is correlated with the input, hence a goal of 100% accuracy is unrealistic and unreachable. Nevertheless, through the proposed statistical framework we were able to draw interesting results. Because of the heterogenicity of the term variables that reached significance, it is reasonable to admit that our forecast logistic regression equations may differ in larger and better-balanced data studies. Nevertheless, our findings prove that some of the chosen variables are indeed strongly and significantly correlated with the delivery mode, and that the chosen artificial intelligence technique is appropriate for solving the real-life problem of predicting the mode of delivery. Our results could be helpful for counseling and aiding in decision-making for patients who have an otherwise elevated risk for cesarean delivery at baseline.
      Figure 6
      Figure 6Looping mechanism for F1-score for optimizing the threshold of the significant variables for weeks 37 to 40
      Iliescu. Statistical framework for labor outcome prediction. Am J Obstet Gynecol MFM 2022.
      A, PD for week 37, B, BMI for week 37, C, PD for week 38, D, BMI for week 38, E, age for week 39, F, BMI for week 39, G, EFW for week 40, H, OP for week 40, and I, BMI for week 40.
      BMI, body mass index; EFW, estimated fetal weight, OP, occiput posterior; PD, progression distance.
      Figure 6
      Figure 6Looping mechanism for F1-score for optimizing the threshold of the significant variables for weeks 37 to 40
      Iliescu. Statistical framework for labor outcome prediction. Am J Obstet Gynecol MFM 2022.
      A, PD for week 37, B, BMI for week 37, C, PD for week 38, D, BMI for week 38, E, age for week 39, F, BMI for week 39, G, EFW for week 40, H, OP for week 40, and I, BMI for week 40.
      BMI, body mass index; EFW, estimated fetal weight, OP, occiput posterior; PD, progression distance.
      Table 8Cutoff values and thresholds
      Iliescu. Statistical framework for labor outcome prediction. Am J Obstet Gynecol MFM 2022.
      Variable/wkMethodThreshold value/method valueCutoff value
      PD/37ROC/G-mean0.7724/0.68140.74
      ROC/Youden's J0.7709/0.668
      Precision recall/F1-score0.7721−/0.5098
      BMI/37ROC/G-mean0.2332/0.648125.07
      ROC/Youden's J0.2330/0.6480
      Precision recall/F1-score0.2327/0.5283
      PD/38ROC/G-mean0.0029/0.6481−0.24
      ROC/Youden's J0.0028/0.6479
      Precision recall/F1-score0.0000/0.396
      BMI/38ROC/G-mean0.0032/0.655125.07
      ROC/Youden's J0.0034/0.6579
      Precision recall/F1-score0.2364/0.5333
      Age/39ROC/G-mean0.4526/0.553328
      ROC/Youden's J0.3484/0.5533
      Precision recall/F1-score0.0000/0.6667
      BMI/39ROC/G-mean0.2722/0.669927.04
      ROC/Youden's J0.2484/0.5533
      Precision recall/F1-score0.2949/0.5333
      EFW/40ROC/G-mean0.541/0.6623490
      ROC/Youden's J0.541/0.661
      Precision recall/F1-score0.412/0.5
      BMI/40ROC/G-mean0.3720/0.679926.5
      ROC/Youden's J0.3481/0.5633
      Precision recall/F1-score0.3942/0.5433
      OP/40ROC/G-mean0.731/0.6671
      ROC/Youden's J0.541/0.662
      Precision recall/F1-score0.4121/0.5
      BMI, body mass index; EFW, estimated fetal weight; OP, occiput posterior; PD, progression distance; ROC, receiver operating characteristic.
      The protocol involves the use of TPU, and measurement of several variables that proved significant for the delivery mode. Because these measurements are not widely used in general practice, it may be argued that the applicability to real-life clinical care could be a concern. It could be considerably challenging to provide training for physicians and/or sonographers in order for this protocol to be part of the standard clinical evaluation. However, practice guidelines for this technique have been issued years ago on the basis of a considerable body of literature.
      • Iliescu DG
      • Adam G
      • Tudorache S
      • Antsaklis P
      • Cernea N
      Reply: To PMID 22302748.
      Moreover, it was previously reported that the transperineal measurements regarding head situation in the maternal pelvis can be performed at high stations, before labor onset.
      • Levy R
      • Zaks S
      • Ben-Arie A
      • Perlman S
      • Hagay Z
      • Vaisbuch E.
      Can angle of progression in pregnant women before onset of labor predict mode of delivery?.
      ,
      • Dietz HP
      • Lanzarone V.
      Measuring engagement of the fetal head: validity and reproducibility of a new ultrasound technique.
      ,
      • Eggebø TM
      • Gjessing LK
      • Heien C
      • et al.
      Prediction of labor and delivery by transperineal ultrasound in pregnancies with prelabor rupture of membranes at term.
      What other variables could be added to the term prediction of labor outcome? Rizzo et al
      • Ghi T
      • Eggebø T
      • Lees C
      • et al.
      ISUOG Practice Guidelines: intrapartum ultrasound.
      proposed US assessment of the fetal HC and maternal subpubic angle obtained from a reconstructed coronal plane on 3D US performed translabially at 36 to 38 weeks of gestation.
      • Iliescu DG
      • Adam G
      • Tudorache S
      • Antsaklis P
      • Cernea N.
      Quantification of fetal head direction using transperineal ultrasound: an easier approach.
      In this pilot study, we attempted to involve as many significant variables of the labor mechanism as possible, but we also had to take into account the general limitations of the maternity centers; therefore, we did not consider 3D acquisitions and interpretations.

      Strengths and limitations

      Strengths

      Prediction of the delivery mode certainly represents an issue of great interest within obstetrics and gynecology. Although the predictors presented are not particularly new and the findings are not practice-changing, our pilot study proposed an original approach to this issue. This observational prospective cohort study of unselected low-risk primigravidae dynamically assessed fetal and maternal clinical and US features, previously proposed for labor outcome predictions. We adopted a wide range of variables to obtain optimal logistic regression predictions, and to investigate the role of each parameter. In addition, we used a wide range of statistical methods to find reliable and valid predictions for labor outcome.
      Thus, our research serves well to provide a framework to which more data could be added for more accurate predictions.

      Limitations

      Although the study size surpassed those of similar previous research, it remained low because the dataset was unbalanced given the discrepancy between the vaginal and cesarean numbers of births. This was owing to the enrollment design of the study: unselected/consecutive term primiparous. However, because of this limitation, the model was sensitive to false-positives. Secondly, although the overall sample size achieved 95% power, when dividing the data per weeks, the sample size decreased, hence the need for a larger population to increase the power analysis. However, this was only a pilot study, and our group continues enrollment to achieve a significant number of patients delivered vaginally and by cesarean delivery at each gestational age.
      Measurement of the EFW variable implied some limitations at the gestational age proposed for the model given that some fetal biometry measurements (biparietal diameter, HC, abdominal circumference) are reportedly less accurate with advancing gestational age.
      This pilot study included a population of women with significantly lower BMIs compared with other populations, which could affect the generalizability of the study.

      Conclusions

      PD and BMI are significantly and strongly correlated with the delivery mode at 37 and 38 weeks and apparently can be used to forecast the delivery mode when the pregnancy reaches term. BMI is predictive of labor outcome throughout term evaluations.
      There are strong and significant correlations among the “head descent” US measurements (AOP, DA, PD, and HPD) at term, before labor onset.
      EFW and PD measurements can be used to forecast the delivery mode in the week before delivery, perhaps as part of a policy for pregnant women with prelabor clinical signs. In addition, HPD and AOP variables had almost significant P levels of.055 and.07, respectively, during the week before delivery.
      Larger studies with more data, particularly better-balanced data, are needed.

      References

        • Dietz HP
        • Lanzarone V
        • Simpson JM.
        Predicting operative delivery.
        Ultrasound Obstet Gynecol. 2006; 27: 409-415
        • Levy R
        • Zaks S
        • Ben-Arie A
        • Perlman S
        • Hagay Z
        • Vaisbuch E.
        Can angle of progression in pregnant women before onset of labor predict mode of delivery?.
        Ultrasound Obstet Gynecol. 2012; 40: 332-337
        • Gillor M
        • Vaisbuch E
        • Zaks S
        • Barak O
        • Hagay Z
        • Levy R.
        Transperineal sonographic assessment of angle of progression as a predictor of successful vaginal delivery following induction of labor.
        Ultrasound Obstet Gynecol. 2017; 49: 240-245
        • Dietz HP
        • Moore KH
        • Steensma AB.
        Antenatal pelvic organ mobility is associated with delivery mode.
        Aust N Z J Obstet Gynaecol. 2003; 43: 70-74
        • Rane SM
        • Guirgis RR
        • Higgins B
        • Nicolaides KH.
        Pre-induction sonographic measurement of cervical length in prolonged pregnancy: the effect of parity in the prediction of the need for Cesarean section.
        Ultrasound Obstet Gynecol. 2003; 22: 45-48
        • Roman H
        • Verspyck E
        • Vercoustre L
        • et al.
        Does ultrasound examination when the cervix is unfavorable improve the prediction of failed labor induction?.
        Ultrasound Obstet Gynecol. 2004; 23: 357-362
        • Rane SM
        • Guirgis RR
        • Higgins B
        • Nicolaides KH.
        The value of ultrasound in the prediction of successful induction of labor.
        Ultrasound Obstet Gynecol. 2004; 24: 538-549
        • Rane SM
        • Guirgis RR
        • Higgins B
        • Nicolaides KH.
        Models for the prediction of successful induction of labor based on pre-induction sonographic measurement of cervical length.
        J Matern Fetal Neonatal Med. 2005; 17: 315-322
        • Dietz HP
        • Lanzarone V.
        Measuring engagement of the fetal head: validity and reproducibility of a new ultrasound technique.
        Ultrasound Obstet Gynecol. 2005; 25: 165-168
        • Eggebø TM
        • Gjessing LK
        • Heien C
        • et al.
        Prediction of labor and delivery by transperineal ultrasound in pregnancies with prelabor rupture of membranes at term.
        Ultrasound Obstet Gynecol. 2006; 27: 387-391
        • Dupuis O
        • Silveira R
        • Zentner A
        • et al.
        Birth simulator: reliability of transvaginal assessment of fetal head station as defined by the American College of Obstetricians and Gynecologists classification.
        Am J Obstet Gynecol. 2005; 192: 868-874
        • Oláh KS.
        Reversal of the decision for caesarean section in the second stage of labour on the basis of consultant vaginal assessment.
        J Obstet Gynaecol. 2005; 25: 115-116
        • Buchmann E
        • Libhaber E.
        Interobserver agreement in intrapartum estimation of fetal head station.
        Int J Gynaecol Obstet. 2008; 101: 285-289
        • Wolpert DH
        • Macready WG.
        No free lunch theorems for optimization.
        IEEE Trans Evol Computat. 1997; 1: 67-82
        • Barbera AF
        • Pombar X
        • Perugino G
        • Lezotte DC
        • Hobbins JC.
        A new method to assess fetal head descent in labor with transperineal ultrasound.
        Ultrasound Obstet Gynecol. 2009; 33: 313-319
        • Kalache KD
        • Dückelmann AM
        • Michaelis SA
        • Lange J
        • Cichon G
        • Dudenhausen JW.
        Transperineal ultrasound imaging in prolonged second stage of labor with occipitoanterior presenting fetuses: how well does the ‘angle of progression’ predict the mode of delivery?.
        Ultrasound Obstet Gynecol. 2009; 33: 326-330
        • Molina FS
        • Terra R
        • Carrillo MP
        • Puertas A
        • Nicolaides KH.
        What is the most reliable ultrasound parameter for assessment of fetal head descent?.
        Ultrasound Obstet Gynecol. 2010; 36: 493-499
        • Ramanathan G
        • Yu C
        • Osei E
        • Nicolaides KH.
        Ultrasound examination at 37 weeks’ gestation in the prediction of pregnancy outcome: the value of cervical assessment.
        Ultrasound Obstet Gynecol. 2003; 22: 598-603
        • Tutschek B
        • Braun T
        • Chantraine F
        • Henrich W.
        A study of progress of labour using intrapartum translabial ultrasound, assessing head station, direction, and angle of descent.
        BJOG. 2011; 118: 62-69
        • Debby A
        • Rotmensch S
        • Girtler O
        • Sadan O
        • Golan A
        • Glezerman M.
        Clinical significance of the floating fetal head in nulliparous women in labor.
        J Reprod Med. 2003; 48: 37-40
        • Shin KS
        • Brubaker KL
        • Ackerson LM.
        Risk of cesarean delivery in nulliparous women at greater than 41 weeks’ gestational age with an unengaged vertex.
        Am J Obstet Gynecol. 2004; 190: 129-134
        • Oboro VO
        • Tabowei TO
        • Bosah JO.
        Fetal station at the time of labour arrest and risk of caesarean delivery.
        J Obstet Gynaecol. 2005; 25: 20-22
        • Altman DG.
        Practical statistics for medical research.
        Chapman & Hall, London, England1991
        • Hodge JK
        • Klima RE.
        The mathematics of voting and elections: a hands-on approach, Mathematical world series 22.
        American Mathematical Society, Providence, RI2005
      1. Tannenbaum P. Excursions in modern mathematics. 6th ed. Upper Saddle River, NJ: Prentice Hall.

        • Belciug S.
        Artificial Intelligence in Cancer-diagnostic to tailored treatment.
        Elsevier Academic press, San Diego, USA2020
        • Rizzo G
        • Aiello E
        • Bosi C
        • D'Antonio F
        • Arduini D
        Fetal head circumference and subpubic angle are independent risk factors for unplanned cesarean and operative delivery.
        Acta Obstet Gynecol Scand. 2017; 96: 1006-1011
        • Ghi T
        • Eggebø T
        • Lees C
        • et al.
        ISUOG Practice Guidelines: intrapartum ultrasound.
        Ultrasound Obstet Gynecol. 2018; 52: 128-139
        • Iliescu DG
        • Adam G
        • Tudorache S
        • Antsaklis P
        • Cernea N.
        Quantification of fetal head direction using transperineal ultrasound: an easier approach.
        Ultrasound Obstet Gynecol. 2012; 40: 607-608
        • Iliescu DG
        • Adam G
        • Tudorache S
        • Antsaklis P
        • Cernea N
        Reply: To PMID 22302748.
        Ultrasound Obstet Gynecol. 2013; 41: 100-101