AI quickly analyzes pregnancy data, outpacing human teams in healthcare research

A team of researchers from the University of California San Francisco and Wayne State University have demonstrated how generative artificial intelligence (AI) can significantly accelerate complex data analysis, particularly in medical research fields like obstetrics. Their findings suggest that AI tools could revolutionize biomedical data science by streamlining processes traditionally handled by human teams.

In a landmark study conducted at a global DREAM challenge, the researchers compiled microbiome data from around 1200 pregnant women across nine separate studies to predict preterm births, aiming to address one of the primary causes of neonatal mortality—premature deliveries. Preterm birth is a leading cause of newborn fatalities worldwide, with approximately 1,000 premature births occurring daily in the United States.

The researchers employed generative AI to process large medical datasets quickly and efficiently. Despite only four out of eight AI chatbots producing usable models, some performed comparably or even surpassed human team achievements. The entire project took six months, a significant reduction from the two years required to consolidate earlier results.

This innovative approach has shown promise for making healthcare research more accessible and efficient. Even a small team, composed of a master’s student and a high school student, was able to develop working prediction models in just minutes. Tasks that previously took experienced programmers days were accomplished swiftly under detailed prompts from the AI.

UCSF Professor Marina Sirota emphasized, “AI tools could help relieve one of the most significant challenges in data science—building our analysis pipelines.” Her observation underscores the transformative potential of generative AI in healthcare research.

Wayne State University’s Adi L. Tarca further elaborated on the benefits: “Generative AI enables researchers to concentrate more on their primary goal, which is understanding scientific questions rather than getting bogged down by coding challenges.”

These findings not only highlight the future possibilities for AI in medical prediction but also offer hope that healthcare research may soon become less resource-intensive and more time-effective.