Garbhini-GA2 could reduce birth risk
CHENNAI: Indian researchers have developed an artificial intelligence (AI)-based model to accurately predict the duration of pregnancy or gestational period of a foetus in the second and third trimester. This model, GARBH-Ini, could reduce errors by as much as three times compared to other existing models programmed in ultrasound devices.
Accurate pregnancy dating is essential for effective antenatal care. However, finding the age of the fetus can be challenging, especially when mothers reach the hospital after the first 12 weeks. But GARBH-Ini, which is based on genetic algorithms, can work wonders on this front, says a new Lancet study.
Backed by the Union government’s Department of Biotechnology, the GARBH-Ini program (Interdisciplinary Group for Advanced Research on BirtH Outcomes DBT India Initiative) is focused on research on birth outcomes, jointly conducted by IIT Madras and Translational Health Science and Technology Institute, Faridabad.
The researchers used genetic algorithm-based methods to develop the Garbhini-GA2 model, which could estimate the pregnancy dates of Indian women in the second and third trimester much better than currently available models.
The model is trained on sample data from a cohort from Gurugram Civil Hospital in Haryana and Safdarjung Hospital, Delhi and independently validated at Christian Medical College in Vellore. Dr Anne George Cherian, professor of obstetrics and gynaecology at Christian Medical College in Vellore, and one of the authors of the Lancet paper, said gestation age plays an important role in predicting the wellbeing of the foetus. “In India, we tend to have a lot of babies that have growth restrictions. We need to deliver them earlier because they don’t tolerate the labour,” she added. She mentioned that many women in India visit the doctor after the first trimester, which makes it hard to predict gestational age. The existing Hadlock formula to calculate the foetal stats are based on samples for white population from North America. The gestation age deduced from the Hadlock formula could differ up to three weeks. “We don’t want to induce a lady without knowing her accurate gestation period,” Dr Anne said.
Dr Himanshu Sinha, coordinator at the Centre for Integrative Biology and Systems Medicine at IIT Madras, who has also led data science for this research, said the model is based on AI/ML works by optimising genetic mutations to reach accuracy. Apart from helping neonatal care, it also helps accurate epidemiological estimates. Dr Sinha said they are using advanced data science and artificial intelligence and machine learning models to predict birth outcomes.
About 30% of pregnant women in India come for their first visit after the first trimester, according to the latest National Family Health Survey.
Elaborating on the model, Dr Sinha said the GARBH-Ini GA-2 is based on genetic algorithms. A genetic algorithm is an optimisation technique inspired by evolution and natural selection principles. “After multiple rounds of mutations, the genetic algorithm comes to a point where no further improvement can be done in the estimation and repeat the process multiple times. Then, we take the top performing formula,” he said.
Dr Sinha said the data for the model comes from the northern part of the country and has been validated at hospitals in the south. The research team plans to further validate the results across different cohorts with tests in different parts of India over two years. It will be set up in Uttar Pradesh, Assam, Puducherry, Maharashtra, Gujarat and a few places in central parts of the country.
He said the model will be improved further and the ultimate end goal is to embed it in ultrasound machines used in the country.