Researchers have developed and tested a new method for building three-dimensional statistical shape models of the foot and ankle that combines weight-bearing computed tomography with magnetic resonance imaging, allowing bone and soft tissue anatomy to be modeled together in a single framework. The methodologic analysis, published in the International Journal of Computer Assisted Radiology and Surgery, reports that the alignment step at the heart of the process proved highly reliable, and that two example models built with the technique performed comparably to previously published shape models.

Statistical shape models are mathematical descriptions of how anatomy varies across a population, constructed from three-dimensional medical images. Within the foot and ankle, weight-bearing computed tomography is widely used to characterize skeletal deformity because it captures the bones while the patient is standing, but it poorly visualizes soft tissue. Magnetic resonance imaging, by contrast, provides improved evaluation of soft tissue structures such as muscles, though it renders bone less accurately. The new process was designed to exploit the strengths of both modalities at once.

The study team collected bilateral weight-bearing CT scans and T1-weighted DIXON MRIs of the lower legs from 14 individuals aged 8 to 18 who had no prior lower limb surgery. The participants belonged to a control group and a hereditary neuropathy group. Bones were segmented from the weight-bearing CT images and then manually aligned into MRI space, so that structures from both imaging types shared a common coordinate framework. Six muscles were segmented semi-automatically from the MRI, and the distal tibia growth plate was segmented manually.

Reliability of the alignment procedure was evaluated using intraclass correlation coefficients, a standard statistical measure of agreement between repeated measurements. The method demonstrated excellent reliability, with an intrarater intraclass correlation coefficient of 1 and an interrater coefficient of 0.999, meaning that the same rater and different raters produced nearly identical alignments when repeating the procedure.

After segmentation, the data underwent cropping, size normalization, and alignment before multi-domain statistical shape models were generated. The researchers evaluated the models in two ways. Qualitatively, they compared the models against expected anatomy. Quantitatively, they calculated three standard metrics: compactness, which reflects how efficiently the model captures variation; generalization, which measures how well the model fits new shapes it has not seen; and specificity, which checks that shapes generated by the model resemble real anatomy.

Two example applications illustrated the method’s flexibility. The first was a combined musculoskeletal model incorporating both bones and segmented muscles. This model demonstrated expected atrophy in neuromuscular disease, consistent with the muscle changes anticipated in the hereditary neuropathy group. The second was a preliminary model of the distal tibia growth plate, which showed variation in metaphyseal thickness across the pediatric cohort.

Both models demonstrated compactness, specificity, and generalization comparable to published statistical shape models. The musculoskeletal model required 11 modes to account for 95 percent of population variance, while the growth plate model required 12 modes. These figures indicate that the combined-modality approach produces models of similar statistical quality to those built from single imaging sources in prior work.

The authors note that the method opens the door to applying shape modeling to novel questions in musculoskeletal morphology by enabling simultaneous modeling of bone and soft tissue anatomy. Because the approach relies on coregistered structures from two imaging modalities, it can capture relationships between skeletal deformity and muscular change that neither modality could reveal alone. The analysis presented examples of muscle and growth plate models, but the researchers describe broader applications across musculoskeletal analysis.

Limitations acknowledged in the analysis include the manual registration step, which links the CT-derived bones to MRI space, and the resolution of magnetic resonance imaging, which constrains the detail available for some structures. The alignment reliability figures of 1 and 0.999 suggest the manual step can be reproduced consistently, but it remains a labor-intensive part of the pipeline. The pediatric age range of the cohort, spanning 8 to 18 years, also reflects the study’s focus on growing skeletons, including the distal tibia growth plate.

For the field of computer-assisted radiology and surgery, the work adds a documented and reliability-tested pipeline for multi-domain shape modeling in a region of the body where weight-bearing geometry matters. By pairing standing CT bone data with MRI soft tissue data in a shared space, the method extends statistical shape modeling beyond bone-only applications, offering a quantitative tool for studying conditions such as hereditary neuropathy and skeletal growth in the foot and ankle.

Subject of Research: A methodologic analysis of combining weight-bearing CT and MRI to generate statistical shape models of foot and ankle musculoskeletal anatomy.

Article Title: Integration of weight-bearing computed tomography and magnetic resonance imaging for advanced musculoskeletal statistical shape modeling of the foot and ankle

Article References: Requist, M. R., Moreno, E., Reddy, C., Peterson, A. C., Mills, M. K., & Lenz, A. L. (2026). Integration of weight-bearing computed tomography and magnetic resonance imaging for advanced musculoskeletal statistical shape modeling of the foot and ankle. International Journal of Computer Assisted Radiology and Surgery. https://doi.org/10.1007/s11548-026-03814-w

Image Credits: AI Generated

DOI: 10.1007/s11548-026-03814-w

Keywords: statistical shape modeling, weight-bearing CT, magnetic resonance imaging, foot and ankle, musculoskeletal imaging, image segmentation, intraclass correlation, hereditary neuropathy, distal tibia growth plate, pediatric imaging, 3D modeling, computer-assisted radiology