A new model reveals the weight of a newborn from routine pregnancy scans
October 17, 2024 | Bharti Dharapuram
Researchers have developed a model to predict foetal growth using the Gompertz formula (bottom left), which can be used to accurately estimate the weight of a baby at birth. The model uses four foetal measurements (panels on the right) from routine ultrasound scans as input. AC: abdominal circumference, FL: femur length, HC: head circumference, and BPD: biparietal diameter. Images from https://doi.org/10.1016/j.medici.2018.01.004 and https://doi.org/10.1007/s13224-021-01574-y shared under a CC by 4.0 license.
Routine ultrasounds of an expectant mother can now be used to predict the weight of her newborn baby using a growth model developed by a recent study. This allows for early interventions during pregnancy as weight deviations linked to risks of neonatal complications and stillbirth can be detected in advance.
Using measurements from at least three routine ultrasound scans across hundreds of pregnant women, the researchers modeled the growth of a foetus over time. They used this to estimate foetal measurements at term, which in turn were used to predict the weight of a baby at birth. Their model is simple and intuitive in capturing foetal growth and more accurate than existing models despite needing lesser data. In the future, this model can be incorporated into ultrasound machines and its predictions used in the clinical assessment of expectant mothers.
The study was carried out by Chandrani Kumari and Rahul Siddharthan of The Institute of Mathematical Sciences (IMSc), Chennai, in collaboration with Uma Ram, obstetrician and gynaecologist at the Seethapathy Clinic and Hospital, Chennai, Leelavati Narlikar from the Indian Institute of Science Education and Research (IISER), Pune, and Gautam Menon from Ashoka University, Sonepat NCR, formerly at IMSc.
During pregnancy, foetal development is monitored at regular checkpoints using ultrasound scans. These measure several linear dimensions of the foetus including abdominal circumference (AC), femur length (FL), head circumference HC), and biparietal diameter (BPD). “Monitoring foetal growth is an integral part of antenatal care because babies that are small or large for their gestational age are at an increased risk of complications,” explains Uma. Currently, foetal growth is assessed using charts, and birth weight can only be predicted using a scan within a week of delivery, which is not usual, she adds.
“We thought, can we model the growth of the foetus as a function of a few parameters,” says Rahul, about the motivation of the study. The search for a suitable mathematical model led them to the Gompertz formula. “It models constrained growth and has previously been used to model tumour growth and foetus volume,” says Chandrani.
The model contains three intuitive parameters, which describe the shape and scale of the relationship between foetal size and age. The researchers found that only the scale parameter needed to be fitted to each foetus, while the shape parameters could be treated as global. The researchers estimated these parameters from ultrasound scans of 774 pregnant women using four biometric measurements (AC, FL, HC, BPD) of a foetus taken at least thrice during pregnancy. Using the estimated scale parameter for an individual foetus in Gompertz formula, the team predicted the foetal size measurements at term. Finally, they used these measurements to train a machine learning model to predict birth weight and verified their results using independent data from 365 pregnant women.
“While growth standards have been published by international organizations, we do not believe all women can be represented by one formula. Taking a step towards personalized medicine, this approach learns one parameter specific to a foetus, enabling us to make a more accurate prediction for that foetus”, says Leelavati.
“We predict birth weight with a much smaller error compared to previous models,” Rahul adds. The team was able to make accurate predictions of birth weight using routine ultrasound scans taken until only 24 and 35 weeks, which is an advantage over existing models, says Chandrani. “Being able to predict the birth weight and knowing if the fetus is maintaining or falling off its growth curve allows us to monitor them better and additionally time delivery,” Uma says.
Leelavati points out that the anonymised data used in the study and the code for predictions are openly available for analysis by others, in line with open science principles. The authors plan to convert their work into an online calculator that can easily be used to predict birth weight. Going forward, they see the possibility of integrating their model into the software of ultrasound scanners to make birth weight predictions readily accessible. As foetal growth patterns can vary between individual mothers, populations, and environments, it would be exciting to extend their research to understand this diversity in the future. They also hope to study other aspects of gestation, gestational complications, and birth outcomes in future projects.
Interdisciplinary meetings and funding for grand challenges can help encourage future collaborations between clinical researchers and academics, Uma says. Availability of open data from large studies and increasing the pool of journals and reviewers with cross-speciality expertise can bolster such efforts, she adds.
Reference: Kumari, C., Menon, G. I., Narlikar, L., Ram, U., & Siddharthan, R. (2024). Accurate birth weight prediction from fetal biometry using the Gompertz model. European Journal of Obstetrics & Gynecology and Reproductive Biology: X, 24, 100344. https://doi.org/10.1016/j.eurox.2024.100344
