AI Body-Composition Mapping on Whole-Body MRI Beats BMI at Predicting Health Risks

Overview

• Researchers report in Radiology that detailed muscle and fat measures from whole-body MRI predict future diabetes, major cardiovascular events, and death more accurately than body mass index.
• A deep-learning pipeline analyzed 66,608 scans from the UK Biobank and the German National Cohort to build age-, sex-, and height-adjusted reference curves and individual z-scores.
• Risk signals were specific: high visceral fat doubled the odds of future diabetes, high intramuscular fat raised the chance of major heart events by about half, and low skeletal muscle increased all-cause mortality.
• The team released an open-source framework and a web-based calculator that lets clinicians benchmark patients, including by extracting body-composition data from existing CT or MRI scans.
• The authors note limits such as a largely White Western European cohort and the rarity of dedicated whole-body MRI, and they plan validation in clinical populations plus disease-specific references, including uses in oncology and monitoring muscle loss during GLP-1 therapy.