National nutrition guidance is becoming much more individualized, thanks in part to ongoing research at the Jean Mayer USDA Human Nutrition Research Center on Aging (HNRCA) at Tufts University—and a School of Engineering faculty member is helping to make it happen.
“The nutrition field is seeing an explosion of data,” said Professor of Electrical and Computer Engineering Eric Miller, “I am interested in using artificial intelligence—or signal processing, or machine learning—to solve problems related to the functioning of the human mind and body.”
The data available to HNRCA experts includes not only details about people’s diets but also longitudinal observations of their cognitive health, Miller explained. And then, focusing on the connection between those two elements, Miller and his team are devising new ways to model and process longitudinal data to infer how diet correlates with—and can be used to predict—cognitive states.
“As an engineer,” Miller said, “what I’ve observed is that nutrition researchers feel that, with all this data we’re collecting, we should be able to make better personalized recommendations. Of particular interest is the subgroup of aging individuals. At the HNRCA, my scientist colleagues are asking, ‘How can diet help them live longer, better lives?’”
As a preliminary step, the researchers are seeking new methods of representing the relevant data in a meaningful way. Currently, nutrition data is reduced to 13 food groups, explained Miller, and numbers assigned to those groups are further collapsed down to a single number, between 0 and 100, called a Healthy Eating Index.
“That one number is supposed to say something about how well you’re eating,” Miller said. But the existence of 13 food groups means that there is more information than can be meaningfully represented in a single number. So before tackling the question of how diet correlates with cognitive state, Miller, collaborating closely with a team of researchers from HNRCA and his Ph.D. student Yang Hu, is working on creating new indices.
“A 13-dimensional space is too big, and a one-dimensional space is restrictive,” observed Miller. “Instead, perhaps there are two or three indices we can produce, to collapse that 13-dimensional vector in order to give us information about nutrition and eating as correlated with cognition, in ways we can actually interpret.”
The hope is to define the ways in which diet predicts cognitive health and then stave off declines in cognition through the use of precision nutrition—diet plans optimized for individuals or subgroups.