Predicting Obesity Rates in Missouri Using Satellite Imagery and Deep Learning
Author Information
Author(s): Xiao Cao, Maharan Adashya, Asteris Panagiotis G, Gangavarapu Haritha, Dahu Butros M
Primary Institution: University of Missouri Institute for Data Science and Informatics
Hypothesis
Can deep convolutional neural networks applied to satellite imagery effectively predict obesity prevalence in Missouri?
Conclusion
The study demonstrates that integrating deep learning and spatial modeling can accurately predict obesity prevalence based on environmental features from satellite imagery.
Supporting Evidence
- Substantial spatial clustering of obesity rates was found across Missouri.
- The spatial lag model demonstrated strong predictive performance, with an R2 of 0.93.
- Local indicators revealed regions with distinct high and low clusters of obesity.
Takeaway
Researchers used satellite images and smart computer programs to figure out how many people are overweight in different areas of Missouri, helping to understand where help is needed.
Methodology
The study processed Sentinel-2 satellite images to extract features, merged them with obesity rate data, and used a spatial lag model to predict obesity rates.
Potential Biases
Self-reported height and weight data may introduce bias, leading to underestimation of obesity prevalence.
Limitations
The study's findings may not be generalizable beyond Missouri, and self-reported BMI data may underestimate true obesity rates.
Participant Demographics
The study focused on 1052 census tracts in Missouri, covering a diverse range of urban and rural areas.
Statistical Information
P-Value
0.68
Statistical Significance
p<0.05
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website