Skip to main content
SMFM 2024 Pregnancy Meeting Main banner
Final Program


Icon Legend
This session is not in your schedule.
This session is in your schedule. Click again to remove it.
Presentation Icons
Additional registration fee required
Additional registration fee required
Faculty have requested this content not be shared on social media
Faculty have requested this content not be shared on social media
CME Credit Offered
CME Credit Offered
Poster Icons
Foundation Awardee
Foundation Awardee
No Social Media
No Social Media
Back
Q&A

Ultrasound/Imaging

Poster Session 3

(849) Macrosomia prediction via sonographic and obstetrical parameters using machine learning.

Wednesday, February 14, 2024
8:30 AM – 10:00 AM  
Location: Prince George's Exhibit Hall CDE, Lower Atrium Level
Objective: Assessing the predictive capacity of machine learning (ML) algorithm, incorporating sonographic and obstetrical parameters, in predicting macrosomia

Study Design: Retrospective cohort study including all parturients at a single tertiary university who had a singleton live birth at ≥ 36 gestational weeks between July 2012 and September 2020, in whom a sonographic estimated fetal weight (EFW) 14 days prior birth was available. The study employed a total of 39 features, comprising biometry sonographic measurements such as abdominal circumference (AC), femur length (FL), bi-parietal diameter (BPD), and head circumference (HC), along with maternal data encompassing demographic information, anthropometric measurements, obstetrical history, chronic and gestational morbidity, and glucose values for gestational diabetes screeaining and diagnosis. Gradient boosting-catboost was utilized to estimate birth weight ≥ 4000 grams. The predictive capacity of the algorithm was then compared to the Hadlock EFW formula accuracy

Results: A total of 20,468 parturients met the inclusion criteria, and 9.29% of the neonates were classified as macrosomic, defined by a birth weight of ≥ 4000 grams. Notably, the algorithm's performance was influenced most significantly, in descending order, by AC, FL, BPD, fetal gender, HC, gestational age, and parity. The algorithm's predictive capacity for macrosomia exhibited an area under the curve (AUC) of 0.93 (95% CI 0.927-0.936), outperforming the AUC of 0.89 observed for the Hadlock’s EFW formula (Figure 1). When employing Youden's index, the algorithm demonstrated a positive predictive value (PPV) of 0.32 and a negative predictive value (NPV) of 0.99.

Conclusion:

The predictive algorithm demonstrated superior accuracy compared to the traditional EFW formula in predicting macrosomia, resulting in a remarkably high NPV. The incorporation ML algorithm into clinical care has the potential to greatly assist in patient counseling regarding the optimal timing and mode of delivery, thereby potentially reducing neonatal morbidity risks.