In an effort to understand obesity on a grand scale, AI is now being trained to analyse the urban factors via imagery on Google Maps.
On the surface, the following sounds like a schoolyard jab aimed at the expanding hips of the woman who birthed you. Your mama is so obese, computers can spot her from space.
But, it seems that science is doing exactly that (AI analysis of obesity, not mum jokes), as they’ve trained a complicated matrix to figure out how obese your suburb is according to numerous geographical factors, i.e. the number of parks v the number of chicken places v the number of hills.
“We propose a method for comprehensively assessing the association between adult obesity prevalence and the built environment that involves extracting neighbourhood physical features from high-resolution satellite imagery,” the team explains in a new paper.
The researchers, from the University of Washington, fed some 150,000 high-resolution Google Maps images into a convolutional neural network (CNN) – a type of AI that uses deep learning to independently analyse and identify patterns.
It’s not 100% accurate, but the AI figured to about 65% to what the estimates we humans have cobbled together with far more articulate data. The researchers are confident their system could offer an easy blanket tool to help studies gauging obesity risk.
“Our approach consistently presents a strong association between obesity prevalence and the built environment indicator across all four regions, despite varying city and neighbourhood values,” the authors explain.
I mean, the findings seem obvious. Socioeconomic factors factor into obesity. We know this. Conversely, open spaces that enable physical activity are a boon for public health. This we know, but allowing a machine to use its superior computational mind to analyse public health on a massive scale opens the research into new possibilities, allowing us to focus on the problem far deeper than the absolute numbers we’re working with now. Example, 63% of Australians are obese, but the reasons why are far more fragmental. One solution does not exist for such a nuanced problem.
Biostatistician Benjamin A. Goldstein from Duke University explains the possibilities further in a critique of the program, stating: “…in the same way a biomarker may serve as a useful indicator of disease risk, these neighbourhood factors can serve as a valuable indicator of health outcomes … going forward, it is likely that machine learning methods will be integral to discovering features associated with disease – likely features never previously suspected.”