CHENNAI: From a Western formula to an India-specific AI model, pregnancy care could be set for a technology shift, with IIT-Madras working towards wider deployment of a home-grown tool to estimate foetal age from ultrasound scans by 2027-28.
Prof Balaraman Ravindran, head, Wadhwani School of Data Science and AI, said that the technology, developed using Indian pregnancy data, would be taken through further validation and the next stage of integration with ultrasound scanning systems.
In an exclusive interaction with DT Next, he added, “The model seeks to address the limitations of applying formulae developed using Western populations to Indian pregnancies. Currently, they use some US formula to determine the age of the foetus. We’ve developed an Indian babies’ formula.”
The India-specific model, Garbhini-GA2, was developed and externally validated using data from Indian pregnancies. Published research has shown that it significantly reduced errors in estimating gestational age compared with the widely used Hadlock formula. The technology was also externally validated using data from Christian Medical College (CMC), Vellore.
Ravindran stated that the technology had produced “very good results” in work involving CMC Vellore and Faridabad, with the research papers already published. “The focus now is on translating the research into a deployable hospital technology,” he explained. “The technology is all ready. But other than the pilot, we have to figure out whether it can go into the scanning machines centrally in the projects. Researchers are also exploring further work with Army hospitals, as the structured setting could help generate additional evidence, though results from the proposed work were yet to emerge.”
Taking the technology into public health institutions, however, could pose practical challenges. Ravindran opined that the additional time required during ultrasound examinations would have to be considered, particularly in Ghs handling large patient volumes. “The main problem is the time bandwidth. They already spend something like 90 seconds and find it very stressful. So it is a challenge,” he pointed out.
The technology could eventually help improve gestational-age assessment, which is important in monitoring foetal growth and identifying pregnancies that may require closer attention. But wider implementation would depend on validation, workflow considerations and compatibility with existing ultrasound equipment. “By 2027-28, we should try to have something available,” Ravindran told this correspondent.
The foetal-age project is part of the Wadhwani school’s broader expansion into interdisciplinary AI research. Ravindran said that the school plans to strengthen its AI for Science work over the next 2-3 years, while continuing to improve its BTech programme in AI, which was designed from scratch rather than by adding a handful of AI courses to a conventional computer science curriculum.
The school is also working on an AI for Education initiative supported by the Union government. Multiple products are expected to be rolled out over the next year including AI tools that can allow teachers to interact with systems in Indian languages and obtain assistance with teaching, exercises and practice. “Some of the models are being designed to operate on mobile devices with limited dependence on cloud connectivity,” Ravindran explained. “IIT-M is also pursuing AI applications in healthcare through work involving its medical sciences and technology, biotechnology and biomedical engineering departments.”