Radiology. 2026 May;319(2):e251939. doi: 10.1148/radiol.251939.
ABSTRACT
Background Body composition (BC) is associated with cardiometabolic risk. However, using BC to predict future disease risk is challenging, as it may reflect body size or age instead of poor health. Purpose To calculate age-, sex-, and height-normalized BC metrics from MRI scans in over 66 000 individuals from the general population and to assess the prognostic value of these metrics for cardiometabolic outcomes beyond traditional risk factors. Materials and Methods In this retrospective study, age-, sex-, and height-specific BC z-scores derived from whole-body MRI scans were calculated using an open-source fully automated deep learning framework. Data were sourced from the UK Biobank (UKB) and German National Cohort between April 2014 and May 2022, including subcutaneous adipose tissue, visceral adipose tissue (VAT), skeletal muscle (SM), SM fat fraction, and intramuscular adipose tissue (IMAT) to provide an open-source web-based z-score calculator, evaluated against reference-standard radiologist labels. Multivariable Cox regression was used to assess the prognostic value of z-score categories (low: z < -1; middle: z = -1 to 1; high: z > 1) for incident diabetes, major adverse cardiovascular events, and all-cause mortality beyond cardiometabolic risk factors in the UKB. Results Age-, sex-, and height-specific BC z-scores were calculated using data from 66 608 individuals (mean age, 57.7 years ± 12.9 [SD]; 34 443 male; mean body mass index [calculated as weight in kilograms divided by height in meters squared], 26.2 ± 4.5). In multivariable-adjusted Cox regression, z-score risk categories had hazard ratios of up to 2.26 for incident diabetes (high VAT category), 1.54 for incident major adverse cardiovascular events (high IMAT), and 1.44 for all-cause mortality (low SM) compared with middle categories. Conclusion Whole-body MRI-derived BC z-scores were used to identify at-risk individuals and predict cardiometabolic outcomes and mortality beyond traditional risk factors. An open-source age-, sex-, and height-adjusted z-score calculator is available at https://circ-ml.github.io. © RSNA, 2026 Supplemental material is available for this article. See also the editorial by Ghosh and Chernyak in this issue.
PMID:42084507 | DOI:10.1148/radiol.251939