Advertisment

AI Could Help Match Donor Milk to Newborns

author-image
Muhammad Jawad
Updated On
New Update
AI Could Help Match Donor Milk to Newborns

A team of researchers led by Timothy Chan is leveraging machine learning to optimize the macronutrient content of pooled human donor milk recipes. Their research is featured in a new publication in Manufacturing and Systems Operations Management and was conducted in collaboration with Mount Sinai Hospital’s Rogers Hixon Ontario Human Milk Bank, which lends support to preterm and sick babies across Ontario.

Advertisment

The Problem with Current Milk Pooling Methods

At the moment, most milk banks, including Mount Sinai’s, rely on one-by-one decision making when pooling donor milk. This presents a significant challenge in creating a consistent, nutrient-rich milk product for sick and premature babies in neonatal intensive care units. And while there are studies that reveal milk from donors early in their postpartum period tends to be protein-rich, a precise prediction of macronutrient content would allow milk bank employees to make more informed decisions.

A Machine Learning Approach to Milk Pooling

Advertisment

The researchers introduced a data-driven framework that bypasses the need for costly devices to analyze donor milk. Instead, they employed an artificial intelligence model to predict the macronutrient content of each donation. This was coupled with an optimization model to increase the consistency of the macronutrient content in the pooled milk product.

An Implementation Trial

A phased study, including a year-long implementation trial, was carried out to validate these AI-informed models. The first phase consisted of collecting data to create the machine learning model, followed by creating an optimization model based on essential macronutrient levels. This method was then validated through a simulation model before being tested in a real-world experiment over 16 months.

Advertisment

Overcoming Challenges

The real-world testing phase saw several obstacles, such as fluctuating donation volumes during the COVID-19 pandemic. The team also had to ensure their AI decisions aligned with the milk bank's operating protocols, which required constant fine-tuning.

Results of this Study

Advertisment

The final phase observed the milk bank for six months and assessed the macronutrient and bacteria levels in pooled recipes. Comparison of optimized recipes with previous ones revealed an increase of up to 75% in reaching protein and fat targets simultaneously, without compromising on safety conditions such as bacteria content. The optimized recipes also took 60% less time to prepare, demonstrating unprecedented operational efficiency.

Benefits for Sick Infants

The optimized recipes also have significant implications for preterm and ill infants. These babies often have underdeveloped digestive systems, which makes it crucial that the milk they consume has a balanced macronutrient profile.

Advertisment

Looking Forward

The team is poised to expand their research scope to include different nutrients in human donor milk and see if these can also be optimized using their models. Ultimately, the team hopes to make their tool applicable to other milk banks. The ideal scenario would be to design a system that can be integrated into hospital systems to optimize recipes sustainably.

The Importance of Donor Mothers and Milk Banks

The team expressed gratitude to the donor mothers and the milk bank staff who made this research possible. Ultimately, the end goal is to see an improved growth and development outcome for the infants who benefit from this optimized donor milk.

Advertisment
Advertisment