Deep Learning Network Could Help Fight Malnutrition in Care Homes — by Watching the Plates

A team of researchers at the University of Waterloo have published a paper detailing a machine learning network, which uses a camera feed of a person's plate to monitor their nutritional intake — as a means of prevent malnutrition in long-term care homes.

“Right now, there is no way to tell whether a resident ate only their protein or only their carbohydrates,” explains Kaylen Pfisterer, PhD, of the problem the team set out to solve. "Our system is linked to recipes at the long-term care home and, using artificial intelligence, keeps track of how much of each food was eaten to make sure residents are meeting their specific nutrient requirements."

With 54 percent of older adults in long-term care suffering from or at risk of malnutrition, it's a major problem to solve — and one which the team has set about fixing via a encoder-decoder food network (EDFN) machine learning network linked to a depth-sensing D-RGB camera system.

Traditionally, nutrition is monitored by staff who record an estimate of a resident's consumption by looking at their finished plate and its contents — an approach that, University Health Network postdoctoral fellow and Waterloo alumnus Robert Amelard claims, has a 50 percent error rate. The team's machine learning system, by contrast, proved accurate to within five percent.

"My vision would be to monitor and leverage any changes in food intake trends as yellow or red flags for the health status of residents more generally and for monitoring infection control," says Pfisterer of the system's potential.

"This system provides improved transparency, approximates human assessors with enhanced objectivity, accuracy, and precision while avoiding hefty semi-automatic method time requirements," the team concludes. "This may help address short-comings currently limiting utility of automated early malnutrition detection in resource-constrained LTC [Long-Term Care] and hospital settings."

The paper describing the work is available in the journal Scientific Reports under open-access terms; data from the project are available from the corresponding author "on reasonable request."