AI-Powered Body Atlas Reveals Flaws in BMI as a Health Measure

A new study has produced a detailed atlas of human body composition using AI and advanced imaging. Researchers analysed whole-body MRI scans from more than 66,000 people to track changes across the lifespan. The findings challenge the reliance on BMI as a key health indicator by revealing more precise risk factors.

The team developed an open-source, fully automated AI framework to process the scans. This system extracts accurate body composition metrics—such as muscle, visceral fat, and intramuscular fat—with minimal human input. It can also be applied to routine chest or abdominal CTs and MRIs, making the approach widely adaptable.

From the data, the study generated reference curves that map how body composition shifts with age. Key findings include a 1.44-fold higher risk of all-cause mortality linked to low skeletal muscle mass. Visceral fat was tied to a 2.26-fold increased risk of diabetes, while high intramuscular fat correlated with a 1.54-fold rise in major cardiovascular events.

The atlas provides a more detailed picture of health risks than BMI alone. By identifying specific body composition markers, it offers clearer insights into mortality, diabetes, and heart disease risks. The open-source AI framework also allows other researchers to build on the findings for future studies.