A large imaging study has put a sharper lens on one of medicine’s most familiar blind spots — body weight does not tell the full story of health risk.
Researchers using automated imaging analysis on whole-body MRI scans from more than 66,000 adults have built what they describe as the most detailed reference map yet of how fat and muscle are distributed across the human body.
The findings, published in Radiology, suggest that muscle quantity, muscle quality, visceral fat, and fat stored within muscle may reveal health risks that body mass index often misses.
The study is likely to attract attention because it challenges a deeply embedded habit in everyday care. For decades, clinicians have used body mass index, or BMI, as a quick way to classify weight status. It is simple. It is cheap. It needs only height and weight. Yet that simplicity is also its weakness.
BMI cannot show whether two people with the same weight have very different bodies beneath the surface. One may carry more skeletal muscle. Another may have more abdominal fat around internal organs. A third may have fat infiltration inside muscle tissue. Those differences matter, particularly for diabetes, cardiovascular disease, cancer care, frailty, and survival.
The new work, carried out by researchers in diagnostic and interventional radiology and based on data from the UK Biobank and the German National Cohort, adds strong evidence that body composition needs a more prominent role in risk assessment.
The research team analysed MRI scans from 66,608 participants who underwent whole-body imaging between April 2014 and May 2022. The average age was 57.7 years. Just over 34,000 participants were male. The average BMI was 26.2, a range often labelled as overweight by conventional categories.
Using an open-source, fully automated deep learning framework, the scientists measured several body composition features from the scans. These included subcutaneous adipose tissue, the fat stored under the skin; visceral adipose tissue, the fat packed around organs in the abdomen; skeletal muscle volume; skeletal muscle fat fraction; and intramuscular adipose tissue, meaning fat found within muscle tissue.
The researchers then adjusted these measurements for age, sex, and height. That step is crucial. A healthy amount of muscle for a tall man in his 30s will not be the same as a healthy amount for a shorter woman in her 70s. Fat distribution also changes with ageing. Men and women tend to show different patterns. Height influences absolute volumes. Without adjustment, comparisons can be misleading.
To solve this, the team expressed each measurement as a z-score. In simple terms, a z-score shows how far a person’s body composition differs from what would be expected for someone of the same age, sex, and height. A score near zero means the person is close to the reference average. A high score means the person is above the expected range. A low score means the person is below it. In this study, low was defined as , middle as to , and high as .
That approach allowed the researchers to ask a more precise question than BMI can answer: is this person carrying more visceral fat, less muscle, or poorer muscle quality than expected for their body type and demographic group?
The answer appeared to have real prognostic value.
High visceral fat was linked with a 2.26x higher risk of future diabetes. High intramuscular fat was associated with a 1.54x higher risk of major adverse cardiovascular events. Low skeletal muscle was associated with a 1.44x higher risk of death from any cause, even after accounting for cardiometabolic risk factors.
These are associations, not proof of direct causation. Still, they are clinically meaningful. They also reinforce a growing view in medicine, excess fat is not the only concern. Muscle matters. Its quality matters too.
That point is especially important because muscle is often discussed only in terms of strength, size, or fitness. The MRI data add another layer. Fat infiltration within muscle can signal poorer muscle quality. A person may appear reasonably built from the outside while carrying more intramuscular fat than expected. Standard body weight cannot show this. BMI cannot show it either. Even some commonly used body composition tools may struggle to capture it with the same anatomical detail.
Researchers involved in the study noted that knowing the volume of intramuscular fat offers a window into muscle quality that conventional measures, including BMI, bioelectrical impedance analysis, and DEXA, cannot easily provide. That distinction could matter in clinical practice, especially when patients are losing weight, recovering from illness, ageing, or undergoing cancer treatment.
The findings also underline the problem with relying too heavily on waist circumference. Waist measurements are useful. They can indicate central adiposity. They are far better than ignoring fat distribution altogether. Yet they still cannot distinguish between visceral fat, subcutaneous fat, muscle mass, or fat inside muscle. Two people with similar waist sizes may have different internal risk profiles.
This is where imaging may change the conversation. Whole-body MRI gives a detailed map of internal tissues without ionising radiation. In this study, artificial intelligence made it possible to process a very large number of scans in a standardised, reproducible way. That scale matters. Manual measurement would be slow, costly, and impractical across tens of thousands of people. Automated analysis turns existing imaging data into something more useful.
The researchers also produced age-, sex-, and height-normalised reference curves for key body composition measures. These reference curves are one of the most practical outcomes of the work. They could help scientists and clinicians judge whether a patient’s fat or muscle profile is typical, unusually high, or unusually low compared with peers.
The team has released an open-source, web-based body composition z-score calculator to support future research and potential clinical translation. The aim is to make datasets more comparable across studies, hospitals, and patient groups. If validated further, such tools could help doctors identify people whose internal body composition places them at higher risk, even when their weight looks unremarkable.
The study may also influence how routine scans are used. The researchers stressed that a dedicated whole-body MRI is not always necessary. If a patient has already had a CT or MRI scan of the chest, abdomen, or pelvis for another reason, body composition information may already be present. It is simply not measured or reported in most routine settings.
That is a powerful idea. Hospitals generate huge volumes of imaging every day. Much of the tissue-level information visible on those scans remains unused. Automated tools could extract and benchmark that information, turning ordinary clinical images into broader health indicators. The approach is often called opportunistic imaging. It means using scans already performed for one reason to gather additional risk information, without asking the patient to undergo a separate test.
The potential uses are broad. In metabolic health, the method may improve early identification of people at risk of diabetes or cardiovascular disease. In older adults, it may help detect low muscle mass or poor muscle quality before frailty becomes obvious. In oncology, it may support treatment planning by identifying patients with low muscle reserves, high fat infiltration, or other body composition patterns linked to poorer outcomes.
Cancer care is a particularly important area for future validation. Patients with reduced muscle mass may face higher risks of treatment toxicity, complications, functional decline, or poorer survival. Body composition measures could help clinicians tailor chemotherapy dosing, nutritional support, rehabilitation, or monitoring. The researchers said next steps include validating the reference curves in clinical populations, especially for predicting treatment toxicity, survival, and recurrence in cancer patients.
The findings may also become relevant in the era of GLP-1 receptor agonists and other weight-loss medicines. These drugs can produce substantial weight reduction. For many patients, that is beneficial. Yet weight loss is not always the same as healthy tissue change. Losing excess visceral fat may improve metabolic health. Losing too much skeletal muscle may not.
A more detailed body composition assessment could help distinguish desirable fat loss from unwanted muscle loss. This is becoming a practical clinical question as more people use medications for obesity and diabetes management.
The scale on its own cannot answer it. Neither can BMI. Imaging-derived muscle and fat metrics might eventually help clinicians monitor whether weight loss is improving internal health, preserving muscle, or creating new risks.
Still, the study should not be read as a call for everyone to undergo MRI scanning. Whole-body MRI is expensive and not universally available. Health systems already face imaging backlogs. The more immediate opportunity may lie in extracting extra information from scans that are already being performed. Wider screening use would require more evidence, careful cost-effectiveness analysis, and clear guidance on what to do with abnormal results.
There are also scientific limits. The study was retrospective. It used large population cohorts, which is a strength, but cohort participants may not perfectly represent every community or patient group. Associations between body composition and outcomes can be influenced by lifestyle, existing disease, medications, ethnicity, socioeconomic factors, and access to care. The researchers adjusted for key factors, but no observational study can eliminate every source of uncertainty.
That said, the message is hard to ignore. Health risk does not sit neatly inside a BMI category. A person with a “normal” BMI may still have high visceral fat or poor muscle quality. Someone labelled “overweight” may have relatively favourable muscle mass. Another person may lose weight during illness, but the clinically important issue may be the loss of muscle rather than fat.
For patients, the practical takeaway is not to dismiss weight, but to put it in context. Body weight is one signal. Waist size is another. Blood pressure, glucose, cholesterol, fitness, diet, sleep, smoking, family history, and medical conditions all matter. Body composition adds a deeper view, especially when measured with reliable tools.
For clinicians, the study offers a route towards more personalised risk assessment. Rather than asking only how heavy a person is, medicine may increasingly ask what is the body composition, what that weight is made of, where fat is stored, how much muscle is present, and whether that muscle is healthy.
BMI will probably remain in use. It is too convenient to disappear. Yet studies like this show why it should not stand alone. As technology is catching up, the future of risk prediction may be less about a single number on a chart, more about a layered profile of fat, muscle, ageing, sex, height, and disease risk.
In that future, a routine scan may reveal far more than the problem it was ordered to investigate. It may reveal the body’s quiet warnings, long before symptoms begin, which makes preventions way effective.
The post Looks Fit, but Hidden Fat May Signal Health Risks, AI Finds first appeared on PP Health Malaysia.
