AI Can Predict Potential Nutrient Deficiencies from Space

AI Can Predict Potential Nutrient Deficiencies from Space thumbnail

New work maps a region’s nutrient landscape

Credit: Thomas Fuchs

Micronutrient deficiencies afflict more than two billion people worldwide, including 340 million children. A lack of vitamins or minerals can have serious consequences for your health. However, it is important to diagnose deficiencies early enough to ensure effective treatment. This requires costly and time-consuming laboratory tests and blood draws.

New research offers a more efficient method. Elizabeth Bondi, a computer scientist at Harvard University, used public satellite data and artificial intelligence to pinpoint areas in which people are most at risk for micronutrient deficiencies. This analysis could lead to early interventions in public health.

Existing AI systems can use satellite information to predict localized issues in food security, but they rely on directly observable characteristics. One example is the ability to estimate agricultural productivity from images of vegetation. It is more difficult to calculate micronutrient availability. Bondi and her colleagues were inspired by research that showed areas near forests have higher dietary diversity. They also wanted to identify other indicators of potential malnourishment. Their work shows that combining data such as vegetation cover, weather and water presence can suggest where populations will lack iron, vitamin B12 or vitamin A.

The team examined raw satellite measurements and consulted with local public health officials, then used AI to sift through the data and pinpoint key features. A food market was one example of a key feature that could be used to predict a community’s risk level. It was inferred from the visible roads and buildings. These features were then linked to specific nutrients that are lacking in four regions of Madagascar. To train and test their AI program, they used real-world biomarker information (blood samples that were tested in labs).

Predictions for regional-level micronutrient deficiencies in populations other than those in the training data sets were accurate to a large extent. They often exceeded estimates based on local public health officials’ surveys. Bondi states that Bondi’s work demonstrates a method that can identify and target vulnerable populations for nutritional support. This may supplement… costly and invasive procedures. The study was detailed at the Association for the Advancement of Artificial Intelligence’s 2022 virtual meeting.

” This is a unique contribution that highlights AI’s potential to advance health,” said Christine Ekenga, Emory University epidemiologist. She was not involved in the study. She adds that it can be difficult to collect health data in low-resource settings due to infrastructure and cost constraints. “The authors have validated a method that can overcome these obstacles

The researchers are developing a software program that allows for this type of analysis to be extended to other countries with public satellite data. Bondi states that the application could be used to allow public health officials to access the information from the system and inform interventions.

This article was originally published with the title “Vitamin Map” in Scientific American 326, 6, 20 (June 2022)

doi: 10. 1038/scientificamerican0622-20a

ABOUT THE AUTHOR(S)

    Rachel Berkowitz is a freelance science writer and a corresponding editor for Physics Magazine. She is bas

    Read More