MiniMAP: An automatic pipeline for paediatric lower limb musculoskeletal model generation
- 1University of Auckland
Abstract
The generation of paediatric musculoskeletal models often requires compromise between data availability, accuracy, and computation time. Statistical shape models allow for rapid generation of bone geometries for use in musculoskeletal modelling. An articulated shape model allows for movement at the joints during optimisation to better fit to experimental data. Here we present a methodology for generating custom paediatric OpenSim models using an articulated shape model of the lower limb generated from CT scans of over 300 children aged 4 to 18 years old. Bone geometries are predicted using demographic factors (age, height, mass, and sex), linear bone measurements (bone widths and lengths), and 12 skin surface landmarks (L/R ASIS, L/R PSIS, medial/lateral epicondyles, and medial/lateral malleoli). To account for skin padding (bone to skin surface distance) a random forest algorithm was developed for skin padding prediction for these 12 landmarks. Accuracy of the model prediction was assessed using a leave-one-out analysis to assess landmark errors, bone surface errors, and clinical bone measurement errors (linear measurements and angular values). Motion capture and MRI data were collected from two paediatric test cases to test the workflow on unseen data. Skin padding values varied widely in this population with average values from 3.4mm at the lateral malleolus to 25.9mm at the PSIS. Skin padding was best predicted using age, height, mass, ASIS width, epicondylar width, and malleolar width with an average RMSE of 4.68mm. The articulated shape model predicted accurate bone geometries from skin surface landmarks with average bone surface RMSEs of 2.64 {+/-} 0.99mm, 1.91 {+/-} 0.59mm, and 1.69 {+/-} 0.56mm in the pelvis, femora, and tibiae/fibulae respectively. In the two unseen test cases, average bone surface errors ranged from 1.64 {+/-} 1.2mm to 3.41 {+/-} 3.1mm. This study demonstrated the accuracy of "MiniMAP", for the rapid generation of custom paediatric musculoskeletal models.
Related
- Automatic Generation of Expert-Level Neuron Segmentation Masks from Fluorescence Microscopy Images for Non-Invasive Deep Learning Analysis of Phase-Contrast Images
- Beyond Binary Rewrite Decisions: Separating Rewrite-Intensity Control from Level Prediction in a Frozen-LLM Generation-and-Selection Pipeline for Korean Easy-Read Drafting
- Automatic parcellation of human cortical gyri and sulci using standard anatomical nomenclature
- The (implicit) force is with you: an application of the implicit function theorem for automatic protein force fields reparametrization
- Comment on "The utility of magnetic resonance imaging of the spinal cord in paediatric transverse myelitis: A systematic review
BibTeX
@misc{carman2026minimap,
title = {{MiniMAP: An automatic pipeline for paediatric lower limb musculoskeletal model generation}},
author = {L. Carman and T. Salemink Waldren and T. Besier and J. Choisne},
howpublished = {bioRxiv},
year = {2026},
month = oct,
doi = {10.64898/2026.10.26.761767},
url = {https://www.biorxiv.org/content/10.64898/2026.10.26.761767v1},
note = {CC BY},
}
RIS
TY - UNPB
TI - MiniMAP: An automatic pipeline for paediatric lower limb musculoskeletal model generation
AU - L. Carman
AU - T. Salemink Waldren
AU - T. Besier
AU - J. Choisne
PY - 2026
DA - 2026/10/26
PB - bioRxiv
DO - 10.64898/2026.10.26.761767
UR - https://www.biorxiv.org/content/10.64898/2026.10.26.761767v1
AB - The generation of paediatric musculoskeletal models often requires compromise between data availability, accuracy, and computation time. Statistical shape models allow for rapid generation of bone geometries for use in musculoskeletal modelling. An articulated shape model allows for movement at the joints during optimisation to better fit to experimental data. Here we present a methodology for generating custom paediatric OpenSim models using an articulated shape model of the lower limb generated from CT scans of over 300 children aged 4 to 18 years old. Bone geometries are predicted using demographic factors (age, height, mass, and sex), linear bone measurements (bone widths and lengths), and 12 skin surface landmarks (L/R ASIS, L/R PSIS, medial/lateral epicondyles, and medial/lateral malleoli). To account for skin padding (bone to skin surface distance) a random forest algorithm was developed for skin padding prediction for these 12 landmarks. Accuracy of the model prediction was assessed using a leave-one-out analysis to assess landmark errors, bone surface errors, and clinical bone measurement errors (linear measurements and angular values). Motion capture and MRI data were collected from two paediatric test cases to test the workflow on unseen data. Skin padding values varied widely in this population with average values from 3.4mm at the lateral malleolus to 25.9mm at the PSIS. Skin padding was best predicted using age, height, mass, ASIS width, epicondylar width, and malleolar width with an average RMSE of 4.68mm. The articulated shape model predicted accurate bone geometries from skin surface landmarks with average bone surface RMSEs of 2.64 {+/-} 0.99mm, 1.91 {+/-} 0.59mm, and 1.69 {+/-} 0.56mm in the pelvis, femora, and tibiae/fibulae respectively. In the two unseen test cases, average bone surface errors ranged from 1.64 {+/-} 1.2mm to 3.41 {+/-} 3.1mm. This study demonstrated the accuracy of "MiniMAP", for the rapid generation of custom paediatric musculoskeletal models.
ER -