The information is there. We just don't routinely measure it.
For generations, medicine has reduced the complexity of the human body to a single ratio of weight and height — a number that says nothing of what lies beneath the skin. A team of German researchers has now trained artificial intelligence on more than 66,000 MRI scans to map the true architecture of fat and muscle, revealing that where fat lives and how deeply it infiltrates muscle tissue predicts diabetes, heart events, and death far more faithfully than BMI ever could. The work arrives not as a distant promise but as an open-source tool that can read body composition from scans hospitals are already taking every day, quietly redefining what it means to measure a person's health.
- BMI has long been medicine's blunt instrument for risk assessment, yet it cannot distinguish between a pound of fat wrapped around an organ and a pound of muscle holding a body upright.
- Intramuscular fat — fat that infiltrates muscle tissue itself — raised the risk of major cardiovascular events by 54 percent, a danger invisible to any scale or tape measure.
- Low skeletal muscle mass carried a 44 percent higher risk of death from any cause, independent of other risk factors, meaning the quality of what holds us up matters as much as how much of it remains.
- The AI system normalizes measurements by age, sex, and height, giving clinicians the reference standards they have never had — converting raw data into a clear picture of how far any individual deviates from healthy peers.
- An open-source calculator now allows this analysis to run on routine CT and MRI scans already being taken in hospitals worldwide, removing the barrier between discovery and daily clinical practice.
For decades, doctors have leaned on body mass index as their primary window into metabolic risk — a number that is fast, simple, and, as a growing body of evidence suggests, fundamentally incomplete. A research team in Germany has now built an AI system that reads the hidden architecture of fat and muscle from whole-body MRI scans, and what it reveals challenges the way medicine has long understood health risk.
The study, published in Radiology, drew on more than 66,000 MRI scans collected through the UK Biobank and the German National Cohort between 2014 and 2022. Using deep learning, the researchers mapped precisely how fat and muscle were distributed throughout each body — not merely how much someone weighed relative to their height. The result was the most detailed reference atlas yet of normal body composition across different ages, sexes, and heights.
The findings are striking. Excess visceral fat more than doubled the risk of developing diabetes. Intramuscular fat — the kind that infiltrates muscle tissue itself — raised the risk of major cardiovascular events by 54 percent. And low skeletal muscle mass was associated with a 44 percent higher risk of death from any cause, independent of other cardiometabolic factors. BMI, which cannot distinguish fat from muscle or account for where fat resides, captured none of this.
Senior author Dr. Jakob Weiss, a radiologist at the University Medical Center Freiburg, noted that clinicians have long lacked age- and sex-adjusted reference standards for body composition. His team addressed this by converting measurements into z-scores that show how far any individual deviates from their peers — a tool that makes the data clinically legible for the first time.
Critically, the system does not require dedicated whole-body imaging. Any routine CT or MRI of the chest or abdomen already contains the necessary information; the AI simply extracts and quantifies it. The team released an open-source web calculator to let clinicians and researchers benchmark their own datasets immediately.
Weiss sees the tool reaching into oncology — predicting treatment toxicity and survival — and into the monitoring of patients on weight-loss drugs, where distinguishing healthy fat loss from dangerous muscle loss is essential. The next steps involve validating the reference curves in patient populations and developing disease-specific standards. But the foundation is in place: a way to transform imaging hospitals already perform into a precise, reproducible map of the body's hidden risks.
For decades, doctors have relied on a single number to assess whether you're at risk for heart disease, diabetes, or early death: your body mass index. It's simple, it's fast, and it's almost entirely useless at telling the actual story of what's happening inside your body. A team of researchers in Germany has now built something better—an artificial intelligence system that reads the hidden architecture of fat and muscle in your frame, and what it reveals suggests that the way we've been measuring health risk has been fundamentally incomplete.
The study, published in Radiology, analyzed more than 66,000 whole-body MRI scans collected between 2014 and 2022 through the UK Biobank and the German National Cohort. The participants ranged widely in age, with an average of 57.7 years, and the researchers used deep learning to extract precise measurements of how fat and muscle were distributed throughout each body—not just how much someone weighed relative to their height. What emerged was the most detailed reference map yet of normal body composition across different ages, sexes, and heights.
The findings challenge the primacy of BMI in clinical decision-making. Excess visceral fat—the kind that wraps around your organs—increased the risk of developing diabetes by more than two-fold. Intramuscular fat, the fat that infiltrates muscle tissue itself, raised the risk of major cardiovascular events by 54 percent. But perhaps most striking: low skeletal muscle mass was associated with a 44 percent higher risk of death from any cause, independent of other cardiometabolic risk factors. The quality of your muscle, it turned out, mattered as much as the quantity. BMI, which knows nothing of muscle quality or fat distribution, could not capture any of this.
Dr. Jakob Weiss, the senior author and a radiologist at the University Medical Center Freiburg, noted that clinicians have long lacked reference standards for how body composition changes with age and differs between men and women in healthy people. "BMI does not reliably reflect a person's actual body composition," he said. His colleague Dr. Matthias Jung added that while the medical field has growing evidence that body composition is an independent risk factor for disease and death, these measures are shaped by height and sex and shift substantially over a lifetime. The AI system normalized all of this, converting raw measurements into z-scores that showed how far an individual deviated from their age-, sex-, and height-adjusted peers.
What makes this work immediately practical is that the researchers did not require dedicated whole-body MRI scans. Any routine CT or MRI of the chest or abdomen—imaging that hospitals are already performing every day—contains the information needed. The AI can extract it, measure it, and compare it to the reference standards. The team released an open-source web calculator to allow clinicians and researchers to benchmark their own datasets, removing a major barrier to adoption.
Weiss envisions the tool extending far beyond cardiovascular risk. In oncology, it could help predict treatment toxicity and survival. In patients taking weight-loss drugs like GLP-1 agonists, it could distinguish between healthy fat loss and dangerous muscle loss. "We're already imaging patients every day," Weiss said. "On every scan of the abdomen or chest, the information is there, we just don't routinely measure or report it. AI now allows us to tap into this hidden layer of data in a quantitative, reproducible way."
The next phase involves validating these reference curves in actual patient populations and developing disease-specific standards for different groups. But the foundation is laid: a tool that transforms routine imaging into a window onto the body's composition, revealing risks that BMI has always missed.
Notable Quotes
BMI does not reliably reflect a person's actual body composition.— Dr. Jakob Weiss, University Medical Center Freiburg
It's not only how much muscle you have, but also the quality of that muscle. Intramuscular fat gives us a window into muscle quality that BMI or other methods can't easily provide.— Dr. Matthias Jung, University Medical Center Freiburg
The Hearth Conversation Another angle on the story
Why does muscle quality matter more than just having muscle?
Because muscle can be infiltrated with fat. You can have a large muscle that's weak and metabolically unhealthy. The AI can see that degradation. BMI can't.
So a person with high BMI but good muscle quality might be lower risk than someone with normal BMI and poor muscle?
Exactly. The study shows that skeletal muscle is protective—it's associated with lower mortality. But only if it's actually functional muscle, not marbled with intramuscular fat.
Why haven't clinicians been measuring this before?
Because it required manual analysis of imaging data. You'd have to look at each scan, measure each tissue type by hand. AI automates that. Now it takes seconds.
Can this work on any scan?
Any existing CT or MRI of the chest or abdomen. Hospitals already have thousands of these archived. The AI just extracts what's already there.
What happens next in cancer care?
They can use it to predict which patients will tolerate treatment, which might relapse, and whether someone is losing muscle dangerously during therapy. That's information doctors don't have now.
Is this replacing BMI?
Not yet. But it's showing why BMI was never enough. The question now is whether hospitals will adopt it.