A new artificial intelligence tool developed by engineers at Johns Hopkins University can accurately predict a person’s waist circumference using only basic information like age, height, weight, ethnicity, and education level. This tool could help doctors more effectively assess risks for conditions like diabetes, heart disease, and stroke—risks typically associated with obesity and often evaluated through the body mass index (BMI), a less comprehensive measure.
“Waist circumference is strongly linked to health risks, but it’s not routinely measured in clinics,” the researchers explain. “Our method offers a quicker, more accurate way for doctors to assess obesity risks without needing to measure a patient’s waist directly.”
Alternative approach
Developed by the Artificial Intelligence for Engineering and Medicine Lab at Johns Hopkins, the machine learning tool predicts waist size within a narrow range 95% of the time, offering a reliable alternative to traditional measurements. The researchers believe this AI-driven approach could improve how obesity-related health risks are evaluated in clinical settings.
While BMI remains a standard tool for assessing obesity, it fails to account for factors like body composition, ethnic differences, and age. This means someone with a “normal” BMI could still face significant health risks, while others with a higher BMI might not. In contrast, waist circumference is increasingly recognized as a better predictor of obesity-related conditions. However, it is rarely measured in clinical practice due to the lack of standardized techniques and the time it takes.
To address these challenges, the Johns Hopkins team analyzed data from the National Health and Nutrition Examination Survey (NHANES) and the Look AHEAD (Action for Health in Diabetes) study. They applied a machine learning technique called “conformal prediction” to estimate waist circumference based on patient data, including education level, which serves as a proxy for eating habits. Their model not only provides a waist size prediction but also offers a range of values that reflect the certainty of the estimate.
Outperforming the norm
The researchers highlight that their model outperformed existing AI methods and demonstrated reliability across diverse populations, including those with diabetes. This ability to generalize across different groups is crucial for its use in real-world clinical settings.
A standout feature of the new algorithm is its capacity to measure its own uncertainty, which the researchers say adds an extra layer of trust and accuracy to the predictions. “We didn’t just deliver a single waist circumference estimate—we provided a range that reflects how confident the model is in its prediction,” they explain. This is particularly important in clinical settings, where uncertainty can guide treatment decisions.
Although the results are promising, the team cautions that they are still preliminary. They plan to further test the model in various clinical environments and populations and refine it by including additional factors such as diet and physical activity to make predictions even more precise.
