Modelling anthropometric variability using Human Body Models and full body statistical shape models

Research Project, 2026 – 2028

Virtual testing including Finite Element Human Body Models (FE-HBMs) will soon be included in the rating of vehicle safety. In order to include population variability, and to promote robust safety measures, mesh morphing will be introduced. This is a well-established method that can be used to create HBMs representing various sub populations. The morphing is driven by statistical shape models (SSMs), representing both the skin and skeleton surfaces. Today’s state-of-the-art SSMs have compatibility problems, as different body parts are defined in separate SSMs and when these are combined will not provide a model that is a correct representation of a human and are also subject to US export control. The aim of this project is thus to develop a statistical shape model of a complete and accurately linked human skeleton, based on Swedish medical images, that can be shared freely.

Using well established machine learning (ML) methods, automated segmentation and landmarking will be performed on a medium sized dataset of clinical CT scans from both males and females. Each subject will be described using a common topology (a template mesh of an open-source FE-HBM), enabling dimension reduction (PCA) and regression modelling. The result of this project will be an open dataset, strongly linked to a FE-HBM for evaluation of injury risk in road traffic accidents, describing skeletal variability in the adult population. This dataset when used in combination with the linked FE-HBM enables future projects to evaluate and further develop robust safety systems for the population of road users, including vehicle occupants, powered two-wheel riders and bicyclists.