Big Bones, Big Muscles
A comprehensive MRI analysis reveals skeleton predicts 85% of muscle mass.
Researchers have just published the most comprehensive full-body muscle analysis to date, to uncover exactly what determines how much muscle someone can build. Using artificial intelligence to analyse MRI scans of 102 healthy adults, scientists measured all 70 major muscles and 13 bones from head to toe, finding that your skeletal dimensions (our "frame size”) are far better predictors of muscle mass than height, body weight, or BMI. Total bone volume explained 85% of the variation in total muscle volume, significantly outperforming traditional body size measurements, such as weight (49% of variation) or BMI (essentially zero predictive value when analysed alone). This relationship was observed at both the whole-body level and for individual muscles, where the associated bone volume consistently predicted muscle size more accurately than any other anthropometric measure.
The study also revealed that while males and females differ substantially in muscle distribution, with females carrying relatively more muscle in their lower bodies and males in their upper bodies, both sexes follow the same bone-to-muscle relationship.
These findings suggest that much of what we attribute to genetic differences in “muscle building potential” may actually be driven by skeletal architecture. This has practical implications for understanding strength potential, tracking muscle loss in aging or disease, and setting realistic expectations for physique development.
Aim
This study aimed to create the first comprehensive reference dataset of full-body muscle and bone volumes in healthy adults and determine which factors—body size metrics or skeletal dimensions—best predict individual muscle sizes across the entire body. Researchers used AI to segment all major muscles and bones from whole-body MRI scans, then examine how muscle volume relates to height, weight, BMI, and bone dimensions, with specific attention to differences between males and females. The broader goal was to establish normative values for muscle volume, fat infiltration, and side-to-side asymmetry that could be used in clinical diagnostics, athletic performance assessment, and musculoskeletal modelling.
Methods
The research team recruited 102 active, healthy adults (53 females and 49 males) aged 18–50 years from four imaging sites across the United States and Belgium. Participants had an average age of 30 ± 9 years, height of 1.72 ± 0.11 meters (ranging from 1.52 to 2.01 meters), and body mass of 71.2 ± 11.10 kilograms (ranging from 48.5 to 113.4 kilograms). The distribution of heights and weights was intentionally selected to reflect the US population.
Each participant underwent whole-body MRI scanning using a three-dimensional Dixon sequence that captured both water and fat images. The average scan time was approximately 40 minutes, though this varied from 30 to 50 minutes depending on body size. Scans were acquired in three stages to capture the entire body from the base of the skull through the feet, including detailed coverage of both arms.
The researchers developed an AI algorithm based on a modified 3D U-Net convolutional neural network to automatically segment 140 muscle structures and 23 bone structures (representing 70 unique muscles and 13 unique bones when accounting for bilateral structures). This AI-assisted approach reduced manual segmentation time from approximately 30 hours per subject to under 3 hours, with inference runtime under 15 minutes per scan. All AI-generated segmentations were manually reviewed and refined by trained engineers using 3D Slicer software. The segmentation platform (MuscleView 2.0) received FDA 510(k) clearance, validating its performance across multiple MRI scanner types.
For each segmented muscle, the researchers calculated volume, fat fraction (percentage of fat within the muscle), and asymmetry (difference between left and right sides). They then performed linear regression analyses to determine which predictors—total bone volume, femur volume, height × mass, mass, height, or BMI—best explained variation in total and individual muscle volumes. The analyses were conducted both with sexes combined and separately for males and females to identify sex-specific patterns. Statistical significance was assessed using multiple comparison corrections via the Benjamini-Hochberg false discovery rate method
Book a call with me here.
We’ll sit down and upgrade or build out your entire training and performance program in one session.
Results
Bone Volume: The Dominant Predictor
Total bone volume emerged as the strongest predictor of total muscle volume, explaining 85% of the variance—far exceeding all body size metrics. The product of height and mass came in second at 62% variance explained, followed by height alone at 58%, mass at 49%, and BMI showing essentially no relationship when sexes were pooled. Remarkably, when males and females were analysed together, bone volume-based predictions showed no systematic bias between sexes, whereas all body size predictors consistently overestimated muscle volume in females and underestimated it in males.
Individual Muscle Analysis
At the individual muscle level, the associated bone volume (such as the humerus for shoulder muscles, femur for thigh muscles, tibia for calf muscles) consistently outperformed anthropometric predictors. For 63 of the 70 muscles analysed, bone volume explained significantly more variance in muscle size than height × mass (p < 0.001). Mean r² values across all muscles were: total muscle volume (0.73), associated bone volume (0.67), height × mass (0.51), mass (0.36), height (0.42), and BMI (0.20). The correlations between muscle volume and body size parameters differed significantly between males and females, whereas bone volume correlations showed no significant sex differences.
Muscle Distribution Patterns
The gluteus maximus was the largest individual muscle, accounting for 4.2% of total muscle volume in females and 3.9% in males, while the smallest muscles, like the pronator quadratus, contributed only 0.04% in both sexes. Hierarchical clustering analysis revealed strong sex-specific patterns in muscle distribution: females had relatively greater muscle volume in the lower body (particularly trunk, calf, and deep hip muscles), while males showed greater relative volume in upper body muscles (shoulder, elbow, and wrist). Males demonstrated significantly higher upper-to-lower body muscle volume ratios compared to females (p < 0.001).
Asymmetry and Fat Infiltration
Side-to-side asymmetry was most pronounced in the upper limb, especially forearm muscles, where the supinator, extensor carpi ulnaris, and pronator teres all showed mean asymmetry values exceeding 9%. Fat infiltration varied widely across muscles, with the highest fat fractions found in trunk and scapular muscles: rectus abdominis (15.4% in females, 14.4% in males) and trapezius (17.6% in females, 13.0% in males). Females had higher fat fractions across many muscles, though there was no overall effect of sex on asymmetry magnitude.
Bone-Muscle Coupling Beyond Height
Even after statistically accounting for height, individuals with larger bones relative to their stature had proportionally more muscle. At the whole-body level, total muscle volume adjusted for height strongly correlated with total bone volume adjusted for height. Similar patterns emerged at the individual muscle level, demonstrating that bone and muscle volumes covary beyond their shared relationship with body size.
Key Takeaways
Frame size matters more than you think. Your skeletal dimensions, or what people often call “frame size”, are the single best predictor of how much muscle you can carry, explaining 85% of the variation in total muscle mass across individuals. This suggests that much of what appears to be genetic variation in “muscle-building potential” may actually be driven by skeletal architecture, which is largely genetically determined.
BMI is a poor indicator of muscularity. When analysed properly without pre-adjusting for sex, BMI showed essentially zero relationship with muscle volume. Its apparent predictive power in previous research likely stems from the fact that it separates males from females rather than actually capturing meaningful variation in muscle mass within each sex.
Sex differences are about distribution, not the bone-muscle relationship. While males and females differ substantially in where muscle is distributed (females carry relatively more in the lower body, males in the upper body), both sexes follow the same fundamental relationship between bone size and muscle size. This means that after accounting for skeletal dimensions, male and female muscle sizes are comparable relative to their frames.
Practical applications for lifters and athletes. Understanding that bone volume is the primary determinant of muscle potential has several practical implications: (1) individuals with naturally larger skeletal structures have greater capacity for absolute muscle mass; (2) comparing yourself to others should account for frame size, not just height or weight; (3) strength-to-weight ratios naturally favor those with smaller frames; and (4) realistic physique goals should be calibrated to your skeletal proportions.
Clinical and diagnostic value. The comprehensive reference dataset enables clinicians to assess whether an individual's muscle volume is appropriate for their skeletal size, making it possible to detect muscle loss from aging, injury, disease, or disuse more accurately than current methods. The femur volume alone performed nearly as well as total bone volume in predicting muscle mass, offering a more accessible measurement for clinical settings where full-body imaging isn't feasible.
Asymmetry and fat infiltration norms established. The study provides the first comprehensive reference values for muscle asymmetry and fat content across all major muscle groups, revealing that upper body muscles (especially forearms) show the most asymmetry—likely reflecting hand dominance—while trunk muscles accumulate the most fat. These baseline values can help identify abnormal patterns that may indicate injury, neurological issues, or metabolic dysfunction.
Reference
Riem, L., Pinette, M., DuCharme, O., Pabon, V., Morris, J., Coggins, A., Harold, L., Costanzo, K. E., Cousins, M., Hein, R., Rhodes, M., Lievens, E., Shah, R., Feng, X., Benusa, S., Breeding, T., Nelson, M. D., Derave, W., & Blemker, S. S. (2025). Big bones mean big muscles: An MRI-based dataset of muscle-bone-body size relationships across 70 human muscles of the upper limb, trunk, and lower limb. Journal of Applied Physiology. https://doi.org/JAPPL-00772-2025
You can also find me at dannyleejames.com for stories, personal training insights, and coaching.