AI MRI study links hidden muscle fat to cardiometabolic risk

German researchers used a deep-learning model to analyze paraspinal muscle composition on whole-body MRI and found associations between hidden muscle fat, lean muscle mass, and cardiometabolic risk factors, according to RSNA.

The study, published in Radiology, included 11,348 participants without known preexisting conditions who underwent whole-body MRI at 5 imaging sites. The cohort was 56.9% male, with a median age of 43.

The researchers used a segmentation algorithm to quantify intermuscular adipose tissue and functional muscle tissue in the paraspinal muscles, which run along the spine between the neck and pelvis. RSNA noted that measuring these features previously required time-intensive manual analysis.

Laboratory tests and clinical examinations found previously undiagnosed cardiometabolic risk factors in the study population. These included hypertension in 16.2% of participants, abnormal blood sugar in 8.5%, and unhealthy lipid patterns in 45.9%.

After adjustment for age, sex, physical activity, and study site, higher intermuscular adipose tissue was associated with higher odds of hypertension, abnormal blood sugar, and unhealthy lipid patterns in both sexes.

Higher lean muscle mass was associated with a protective effect against cardiometabolic risk factors only in men, according to the researchers. In women, lean muscle mass remained relatively stable until ages 40 to 50, then declined, a pattern the researchers said overlaps with the menopausal transition.

Low physical activity was also associated with increased intermuscular adipose tissue and decreased lean muscle mass.

Lead researcher Sebastian Ziegelmayer, MD, of the Technical University of Munich, said the study is an initial step toward establishing an imaging-based biomarker for cardiometabolic vulnerability.

Because MRI is already used for many clinical purposes, the method could allow additional risk information to be extracted from scans that are already being performed, RSNA said.

The researchers said further work could extend the analysis to more advanced MRI sequences and assess whether muscle composition reflects broader health status beyond cardiometabolic risk.