The information is already there. We just never measured 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. Now, a team of German researchers has trained an artificial intelligence on more than 66,000 MRI scans to map the true interior landscape of the body: where fat resides, how much muscle endures, and whether that muscle is clean or quietly infiltrated by fat. In doing so, they have not merely improved a diagnostic tool — they have challenged the philosophical premise that a single number can stand in for a life.
- BMI has long functioned as medicine's shorthand for risk, but it cannot distinguish a muscular body from a soft one, nor see the visceral fat coiled silently around organs.
- The AI system reveals that intramuscular fat — fat embedded within muscle tissue itself — raises the risk of major cardiovascular events by 54%, a signal entirely invisible to conventional measures.
- Low skeletal muscle volume correlates with a 44% higher risk of death from any cause, reframing muscle not as an aesthetic concern but as a fundamental survival resource.
- Researchers have released the tool as an open-source web calculator, allowing clinicians to extract body composition data from CT or MRI scans patients are already receiving — no new imaging required.
- The tool's reach extends to oncology and weight-loss drug monitoring, where it can distinguish healthy fat loss from dangerous muscle depletion — a distinction that could determine whether a treatment helps or harms.
For decades, the body mass index has served as medicine's primary shorthand for metabolic risk — a simple calculation of height and weight that, as researchers in Germany have now demonstrated, misses nearly everything that matters. Their AI system, trained on more than 66,000 full-body MRI scans drawn from UK and German cohort studies between 2014 and 2022, measures five distinct body composition metrics and converts each into an age- and sex-adjusted z-score, revealing how far any individual deviates from what is normal for someone like them.
The findings reorder clinical priorities. High visceral fat — the kind that wraps around internal organs — was associated with a 2.26-fold increased risk of diabetes. Intramuscular fat, which infiltrates muscle tissue itself, raised the risk of major cardiovascular events by 54%. Most striking of all, low skeletal muscle volume corresponded to a 44% higher risk of death from any cause, even after controlling for other risk factors. Lead author Matthias Jung noted that muscle quality, not merely quantity, is what the AI uniquely captures — a dimension that BMI, impedance analysis, and even DEXA scanning cannot easily provide.
The researchers have made their system available as an open-source tool, allowing clinicians to extract body composition data from routine CT or MRI scans already being performed — no additional imaging required. The information, as radiologist Jakob Weiss observed, has always been present in the scans. Medicine simply lacked the means to read it systematically.
The applications extend well beyond general risk assessment. In cancer care, the tool could predict which patients will tolerate chemotherapy and which face greater toxicity risk. It also addresses an emerging problem with GLP-1 weight-loss drugs: determining whether a patient losing significant weight is shedding fat or losing the muscle that sustains them. What began as an effort to surpass BMI has opened a far richer window into the body — one that was always there, waiting only to be seen.
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 real story of what's happening inside your body. A team of researchers in Germany has now built something better—an artificial intelligence system trained on more than 66,000 full-body MRI scans that reveals what BMI has always missed: where your fat actually lives, how much muscle you have, and critically, what that muscle is made of.
The study, published in Radiology, analyzed scans from participants in the UK Biobank and the German National Cohort collected between 2014 and 2022. The cohort averaged 57.7 years old, with a mean BMI of 26.2—solidly in the "overweight" category by conventional standards. But BMI, as Jakob Weiss, a radiologist at University Medical Center Freiburg, points out, is a crude instrument. It measures only height and weight. It cannot distinguish between a person who is muscular and a person who is soft. It ignores the fact that where fat accumulates matters enormously—visceral fat, the kind that wraps around your organs, is far more dangerous than subcutaneous fat, the kind under your skin. And it tells you nothing about muscle quality, which turns out to be one of the strongest predictors of whether you'll live a long life or not.
The AI system the researchers built works by automatically analyzing the MRI scans and measuring five key metrics: subcutaneous adipose tissue, visceral adipose tissue, skeletal muscle volume, skeletal muscle fat fraction, and intramuscular adipose tissue. Each measurement is then converted into a z-score—a statistical tool that shows how far an individual deviates from what's normal for someone of their age, sex, and height. This normalization is crucial. A 70-year-old woman naturally has less muscle than a 40-year-old man. The old tools didn't account for that. This one does.
The findings are stark. People with high visceral fat faced a 2.26-fold increased risk of developing diabetes. Those with high intramuscular fat—fat infiltrating the muscle tissue itself—had a 1.54-fold increased risk of major cardiovascular events. And people with low skeletal muscle had a 1.44-fold higher risk of death from any cause, even after accounting for other cardiometabolic risk factors. The last finding is perhaps the most important. It suggests that the amount of muscle you carry is not a vanity metric. It is a survival metric. "It's not only how much muscle you have, but also it's the quality of that muscle," Matthias Jung, the study's lead author, explained. "Knowing the volume of intramuscular fat gives us a window into muscle quality that other methods like BMI, bioelectrical impedance analysis, or DEXA can't easily provide."
The researchers have released their work as an open-source tool—a web-based calculator that clinicians and researchers can use to benchmark their own patients against the reference curves. The implications are immediate and practical. Doctors don't need to order dedicated whole-body MRIs. If a patient is already getting a routine CT scan of the chest or abdomen, the AI can extract the body composition data from that existing image. The information is already there, Weiss notes. Medicine has simply never bothered to measure it systematically before.
The potential applications extend far beyond routine risk assessment. In oncology, the tool could help doctors predict which cancer patients will tolerate chemotherapy well and which are at higher risk of treatment toxicity. It could also solve a problem that has emerged with the rise of weight-loss drugs like GLP-1 agonists: distinguishing between healthy fat loss and dangerous muscle loss. A patient might lose 30 pounds, but if that loss includes significant muscle, their health may actually worsen. The AI can see that distinction instantly.
Weiss and his team are now working to validate these reference curves in actual patient populations and to develop disease-specific versions for other groups. The next frontier is understanding how body composition predicts survival and recurrence in cancer patients. What began as an attempt to improve on BMI has opened a door to a much richer understanding of human health—one that was always visible on the scans doctors were already taking, waiting only for the right tool to make it legible.
Citações Notáveis
BMI does not reliably reflect a person's actual body composition.— Jakob Weiss, M.D., Ph.D., University Medical Center Freiburg
It's not only how much muscle you have, but also it's the quality of that muscle.— Matthias Jung, M.D., lead author of the study
A Conversa do Hearth Outra perspectiva sobre a história
Why does it matter that we can now measure muscle quality? Couldn't doctors always see muscle on an MRI?
They could see it, yes, but they weren't measuring it systematically or comparing it to anything. It's like having a photograph of a forest but never counting the trees. The AI creates a standard—here's what normal muscle looks like for a 55-year-old woman of your height—so you can see whether yours is better or worse.
And the intramuscular fat—the fat inside the muscle itself—that's the surprising finding?
It is. Most people think about fat as something separate from muscle, something you lose or gain. But fat can infiltrate muscle tissue, and that infiltration is a sign of poor muscle quality. It's invisible to BMI. You could be thin by BMI standards and still have this problem.
So someone could have a normal BMI and still be at high risk?
Absolutely. You could be 5'10", weigh 180 pounds, have a BMI of 25—perfectly normal—and have dangerously high visceral fat or low muscle mass. BMI would tell you nothing about it.
The tool is open-source. Why release it freely instead of commercializing it?
Because the researchers want it used. If it's locked behind a paywall or proprietary system, it sits in a lab. Released openly, it becomes part of the standard toolkit. Every hospital can use it on scans they're already taking.
What happens next? Is this going to change how doctors assess risk?
That's the question. The tool exists now. The evidence is strong. But changing medical practice is slow. Doctors still use BMI because it's simple and it's what they learned. This requires a shift in thinking—and access to imaging, which not everyone has.
So it could widen health disparities?
It could, yes. If only wealthy patients with access to MRI get this kind of detailed assessment, while others rely on BMI, you've created a two-tier system. That's a real concern the field will have to grapple with.