A workforce of College of Toronto Engineering researchers, led by Professor Timothy Chan, is leveraging machine studying to optimize the macronutrient content material of pooled human donor milk recipes.
The researchers introduce their data-driven optimization mannequin in a new paper published in Manufacturing and Techniques Operations Administration.
Chan and his workforce labored with Mount Sinai Hospital’s Rogers Hixon Ontario Human Milk Financial institution—which offers donor milk to preterm and sick infants who’re hospitalized throughout Ontario—in addition to Dr. Debbie O’Connor, a professor on the Temerty School of Drugs.
“For quite a lot of causes, many hospitalized infants don’t have a full provide of mom’s milk. On this occasion human donor milk could be lifesaving significantly because it helps to guard preterm infants from necrotizing enterocolitis, a life-threatening bowel illness,” says Dr. Sharon Unger, a neonatologist and the medical director of the Rogers Hixon Ontario Human Milk Financial institution.
“The brand new program developed by Dr. Chan helps to make sure that every batch of human donor milk meets the protein and calorie wants of preterm infants.”
At the moment, many milk banks, together with Mount Sinai’s, depend on particular person determination making when pooling donor milk. This presents a major problem in producing a constant donor milk product that comprises adequate macronutrients for untimely and sick infants in neonatal intensive care items.
“It takes a whole lot of time to create these recipes with no outlined technique,” says Chan.
“Whereas there are research that present that milk that comes from donors who’re early of their postpartum interval tends to be extra protein wealthy, our strategy offers a superb prediction of the particular macronutrient content material that may enable milk financial institution staff to make higher pooling choices.”
Provided that milk banks are sometimes non-profit entities working on lean budgets, a low-cost different to acquiring a constant, nutrient-balanced product could possibly be helpful throughout the whole sector.
Gadgets often called human milk analyzers can be utilized to measure the precise macronutrient content material of every milk pattern at a milk financial institution. Nevertheless, these units are expensive and require in depth regulatory approval to be used, with the consequence that solely half of all milk banks in North America use one. On prime of that, analyzing each donation is a expensive endeavor that’s labor and useful resource intensive.
“Our data-driven framework bypasses the necessity for a tool to investigate the donor milk through the use of a man-made intelligence mannequin to foretell the macronutrient content material of every donation,” says Rachel Wong, a lead researcher of the examine.
“As well as, through the use of an optimization mannequin to decide on which donations to pool collectively, we will enhance the consistency of macronutrient content material within the donor milk product.”
The multi-phased examine included a one-year implementation trial on the Rogers Hixon milk financial institution that was designed to check whether or not AI-informed fashions might assist to fill the hole.
Within the first part, researchers collected the required knowledge to create a machine studying mannequin to foretell the macronutrient content material of the pooled recipes, after which designed an optimization mannequin to create the recipes based mostly on macronutrient necessities, that’s, the required ranges of protein and fats.
The workforce then created a simulation mannequin to check the tactic earlier than embarking on an experiment within the milk financial institution, which passed off over 16 months in 2021 and 2022.
“Since our examine was carried out within the milk financial institution throughout common working hours, reasonably than in a controlled environment, there have been quite a few sudden challenges that we needed to adapt to,” says Wong.
“Throughout the COVID-19 pandemic, the amount of donations fluctuated based mostly on the provincial restrictions—through the lockdown intervals there was an unprecedented enhance within the quantity and quantity of donations.
“We additionally wanted to adapt the AI choices that had already been proposed to make sure that we abided with the milk financial institution’s working protocols.”
The final part of the examine started by observing the milk financial institution’s operation for six months and measuring the fats, protein and micro organism ranges within the pooled recipes.
For the next six months, the milk financial institution used the data-driven optimization framework to create the pooled milk recipes. On the finish of the 12 months, the researchers in contrast the optimized recipes to the earlier ones to evaluate which recipes met the macronutrient targets.
“We discovered that our pooled recipes met the bar for protein and fats concurrently as much as 75% extra usually, with out compromising different elements like a rise in micro organism,” says Chan. “And it took us 60% much less time to make the recipes.”
The workforce’s optimized recipes even have an additional benefit for pre-term and sick infants, who’ve underdeveloped digestive methods that make it particularly essential to make sure that the milk they’re consuming is not overly wealthy in protein or fats.
Chan’s workforce is presently working in direction of increasing this analysis to measure different vitamins in human donor milk to see if their fashions can optimize them. The analysis has gained INFORMS’s 2023 Pierskalla Best Paper Award and an Excellence in High quality and Security award from Sinai Well being.
“Our final purpose is to point out that our device is relevant to different milk banks,” says Chan. “We wish to design a system that may plug into hospital methods to optimize recipes in a manner that’s sustainable for milk financial institution workers.”
Wong says that the whole workforce is grateful to all those that have made the challenge doable.
“We could not have achieved this with out all the moms who donate to the milk financial institution and the workers who work extremely exhausting to offer donor milk to infants throughout Ontario and past,” she says.
“I hope that this analysis will present a framework to assist milk banks throughout North America enhance the consistency of macronutrient content material of their donor milk product. The eventual finish purpose can be to see a downstream influence of improved progress and developmental outcomes for the infants that obtain this donor milk.”
Timothy C. Y. Chan et al, Bought (Optimum) Milk? Pooling Donations in Human Milk Banks with Machine Studying and Optimization, Manufacturing & Service Operations Administration (2023). DOI: 10.1287/msom.2022.0455
University of Toronto
How AI might assist optimize nutrient consistency in donated human breast milk (2023, November 21)
retrieved 21 November 2023
This doc is topic to copyright. Aside from any truthful dealing for the aim of personal examine or analysis, no
half could also be reproduced with out the written permission. The content material is offered for info functions solely.