A routine shoulder MRI may capture only part of the rotator cuff—but this study asks whether that partial view can still reveal whole-muscle size and fat infiltration in three dimensions. Rotator cuff tears remain a challenging clinical problem, causing pain, limited active range of motion, and weakness.

The scale is substantial: rotator cuff repairs are the second most common orthopedic soft tissue surgery, with over two hundred fifty thousand performed annually in the United States. Approximately twenty to fifty percent of individuals over sixty years of age have a known rotator cuff tear.

Despite clinical pervasiveness and the evolution of surgical techniques, outcomes of rotator cuff repair remain highly variable, particularly for larger, retracted, and chronic tears. Muscle atrophy and fat infiltration of the rotator cuff muscles play significant roles in determining healing and functional outcome after repair.

Current assessments rely on subjective grading of a singular computed tomography or magnetic resonance image, including the Goutallier score, Fuchs score, tangent sign, and occupation ratio. These methods have multiple limitations: subjectivity and lack of precision, limited applicability of two-dimensional measurements to the entire muscle structure, bias toward the supraspinatus, and lack of concrete normative comparisons for natural atrophy with age.

Clinical MRI scans often capture only the lateral portion of the rotator cuff muscle structure, so a method must account for variation in capture range to determine three-dimensional muscle volume. The study aimed to develop and validate a volumetric method for all four rotator cuff muscles, their infiltrated fat, and the surrounding bones from clinical MRI scans with typical limited lateral coverage.

It also used healthy clinical MRI scans of varying ages and across sexes to examine age- and sex-related differences in rotator cuff muscle size and fat infiltration. The hypothesis was that rotator cuff muscles exhibit increased atrophy and fat infiltration with age.

Validation in a healthy control population was intended for eventual application in patient or athlete populations. Figure one summarizes the study’s four-step workflow for analyzing clinical rotator cuff MRI scans with partial lateral coverage. The authors first estimate total scapula length, then relate muscle measurements to the proportion of scapula captured, establish how healthy-muscle metrics vary with scapular location, and finally examine sex- and age-related effects using a clinical-scan database.

This matters because it provides a structured way to quantify all four rotator cuff muscles, fat infiltration, and surrounding bones despite limited scan coverage. The dataset consisted of one hundred seventy shoulder MRI scans and forty-seven shoulder CT scans with matching MRI scans, obtained retrospectively from one surgical clinic.

The datasets were separated into a complete-coverage CT group of forty-seven patients and a healthy rotator cuff group of one hundred seventy patients with no diagnosed rotator cuff pathology confirmed by MRI inspection and clinician notes. The MRI scans used a sagittal-plane protocol typical of clinical workflows, with infiltrated fat visualized distinctly from muscle and resolution of at least six millimeters in the sagittal plane and one millimeter in-plane.

For comparisons across age and sex, subgroups were created for male and female biologic sex and six age subgroups spanning fifteen to seventy-nine years. A multivariable linear regression related four MRI measures to total scapula length obtained by CT for thirty-three randomly selected patients.

All variables were normally distributed according to the Shapiro-Wilks test. To evaluate the output equation, predicted scapula length in fourteen scans excluded from the initial regression was compared with the length predicted by the regression.

Figure two shows how the authors connect a partial MRI view of the scapula with complete CT anatomy. Panel A shows scapula segmentation on matched MRI and CT scans, while panel B registers the full CT rendering to the partial MRI rendering. Panel C identifies the MRI’s peak cross-sectional area and measures its position, area, and vertical and horizontal dimensions, then relates these features to total scapula length from CT.

This matters because those measurable landmarks support estimating full scapula length when MRI coverage is incomplete. Segmentation of the rotator cuff musculature, fatty infiltration, and bones from one hundred seventy clinical MRI scans was performed automatically using three-dimensional artificial-intelligence segmentation.

Three trained segmentation engineers vetted the output to ensure accuracy. The regions of interest included the humerus, scapula, clavicle, supraspinatus, infraspinatus, teres minor, subscapularis, and the fat associated with each muscle.

Only intramuscular fat—fat contained within the muscle border—was labeled for each muscle. Figure three summarizes the rotator cuff MRI-processing pipeline, from two-dimensional region-of-interest segmentation and three-dimensional rendering to quantitative analysis.

It shows cumulative scapula, muscle, and intramuscular-fat volumes along medial distance, estimates total scapula length from lateral scapular morphology, and then presents normalized muscle size, relative contribution, and fat infiltration by percentage of scapular depth.

This matters because normalization to scapula length helps account for differences in scan coverage when comparing muscle composition across scans. Each muscle region-of-interest volume was expressed as a function of distance along the scapula.

To account for variation in scan coverage, the full scapula length was determined and each muscle volume was then expressed as a percentage location along the scapula. Muscle and scapula volumes were interpolated between images and expressed by one-percent increments of distance into the scapula.

The clinical scan cohort contained one hundred seventy scans, providing a reference for the range of total coverage obtained. The regression found that sagittal distance of peak cross-sectional area and peak cross-sectional area were significantly correlated with scapula length.

The vertical and horizontal bounds were not significantly correlated with scapula length. The final relationship, with a correlation of zero point ninety-three and a p value below zero point zero zero one, was used to predict scapula length in fourteen test scans.

The average absolute error was two point ninety-two percent, with a standard deviation of one point sixty-eight percent; errors ranged from negative five point seventy-seven to five point seventy-nine percent, with no bias toward under- or over-prediction. Figure four shows how rotator-cuff muscle morphology is represented at different percentages along the scapula.

Panel A pairs anterior and posterior coverage maps with MRI examples from ten, twenty, thirty, and forty percent locations, while panel B summarizes scan coverage across the control database. The chart indicates that approximately ninety percent of the one hundred seventy scans captured at least forty percent of the scapula, whereas less than fifty percent captured fifty percent, supporting the authors’ use of thirty-percent coverage for later analyses.

When partial lateral coverage was compared with full coverage at ten, twenty, thirty, and forty percent, coverage of at least thirty percent consistently achieved a strong significant correlation. At thirty percent coverage, raw-volume correlations with total coverage were high for the scapula, supraspinatus, infraspinatus, teres minor, and subscapularis.

The same pattern occurred for normalized volume and relative contribution across all four rotator cuff muscles at thirty percent. Correlations with total coverage tended to increase as lateral coverage increased. Figure five compares measurements from partial lateral scapula coverage with measurements from complete coverage.

The plots show raw volume, supraspinatus volume across ten, twenty, thirty, and forty percent coverage, normalized volume, and relative contribution for the four rotator-cuff muscles. The authors use these correlations to show that coverage ranges of at least thirty percent produced strong, significant agreement with full-coverage measurements, supporting partial imaging as a way to characterize muscle volume and composition.

Table 2 reports correlations between partial scapula coverage—ten, twenty, thirty, and forty percent—and full-coverage measurements for raw volume, normalized volume, and relative contribution across the rotator cuff muscles. Correlations generally reach strong, significant levels at thirty or forty percent coverage; for example, supraspinatus raw volume is zero point seven one at thirty percent, while teres minor normalized volume is zero point eight two.

This matters because it supports the authors’ statement that at least thirty percent coverage can capture full-coverage muscle characteristics. There was a trend of decreasing rotator cuff size with age for all muscles.

The main effect of age was significant for the supraspinatus, teres minor, and subscapularis. For the supraspinatus, the result was an F statistic of fifteen point sixty-six with a p value below zero point zero zero one and partial eta squared of zero point three three one.

Figure six shows that partial scapula volume is strongly and significantly associated with the volumes of all four rotator cuff muscles when measured between thirty and forty percent along the scapula, with correlation coefficients above zero point six five and p less than zero point zero zero one.

The left panels display these correlations for grouped and sex-split data, while the right panel illustrates the relationship for supraspinatus volume at thirty percent, with participants distinguished by sex and age. This matters because it supports estimating rotator cuff muscle volume from limited-coverage clinical scans.

For fat infiltration, the infraspinatus demonstrated an increase with age and in females. The main effect of sex for the infraspinatus was significant, as was the main effect of age. The sex effect had an F statistic of eight point eighteen and a p value below zero point zero one, while the age effect had an F statistic of three point ninety-one and a p value below zero point zero one.

Figure seven compares males and females across six age groups for each rotator cuff muscle, showing normalized muscle volume, relative contribution, and fat infiltration. The authors report age-related declines in size for the supraspinatus, teres minor, and subscapularis, alongside reduced supraspinatus contribution and increased infraspinatus contribution with age.

Infraspinatus fat infiltration also shows significant effects of both age and sex, highlighting that aging may alter not only muscle quantity, but also its composition and proportional role. At thirty and forty percent lateral scapula coverage, rotator cuff volume, normalized volume, and relative contribution strongly and significantly correlated with total muscle coverage.

The partially captured lateral region from clinical MRI scans can therefore be used to make inferences about total rotator cuff muscle size and relative contribution. A limitation is that tears, increased fatty infiltration, and other features in the medial sixty percent are not captured and cannot be predicted or foreseen.

The relationship between partial and full coverage was valid across a range of muscular pathologies in the CT scans, which included patients preparing for shoulder replacement and patients with deemed healthy rotator cuff musculature. Although the sample size was large at one hundred seventy, subgroup sample sizes were smaller than or equal to the broader cohort once divided across age and sex groups.

Future work should expand and cover larger ranges of patient heterogeneity because healthy clinical rotator cuff scans were limited at the research site. Moderate to strong statistical power was achieved, signifying that sample size was not as much of a concern for interpreting the results.

The method analyzes the entire three-dimensional volume of all rotator cuff muscles and their respective fatty infiltration as a function of location along the scapula using clinical shoulder MRI scans. It accounts for variation in scan coverage, patient size, and patient age because lateral scapula characteristics can accurately predict total scapular length.

In healthy controls of varying age and sex, lateral rotator cuff characteristics showed diminished muscle size with age and increased fatty infiltration in females. Characteristics from the lateral thirty to forty percent of the scapula had a strong, significant relationship with complete coverage.

The methodology can provide a z-score of expected muscle size, relative contribution, and fatty infiltration for patient populations based on the control database. The study shows that lateral coverage of roughly thirty to forty percent of the scapula can strongly track complete rotator-cuff measurements, while age and sex shape specific patterns of muscle loss and fat infiltration.