Virtual testing using Human Body Models (HBMs) is increasingly being introduced in consumer vehicle safety ratings. Organisations such as Euro NCAP, China-NCAP, and the Insurance Institute for Highway Safety are establishing frameworks where HBMs contribute to safety assessment. HBMs have traditionally considered the 50th percentile male, underserving a diverse global population. These developments and current limitations necessitate population-representative anatomical models that capture variability in sex, age, stature, and body composition. Human anatomical variability is commonly represented using statistical shape models (SSMs). These are used to morph generic HBMs such as THUMS, GHBMC, SAFER HBM, and VIVA+ to varied anatomies. However, most existing SSMs are constructed for isolated anatomical regions and later combined to pseudo-full-body models. This approach breaks statistical covariance between anatomical structures and may produce anatomically incoherent full-body geometries. Constructing full-body SSMs requires large datasets of segmented skeletal geometries extracted from medical imaging data. Existing automated segmentation tools, such as TotalSegmentator and Skellytour, exclude distal skeletal regions such as tibia, fibula, radius and ulna and imperfectly generalise to available, local datasets. As a result, current methods do not readily support scalable extraction of complete skeletal geometries suitable for full-body modelling. This project aims to (1) develop an in-house neural network segmentation framework tailored to the available medical imaging datasets and (2) integrate into an end-to-end HBM morphing pipeline. The computationally intensive component of the project is the automated processing of large datasets of CT images to extract full-skeleton geometries, including distal limbs. Following segmentation, skeletal geometries will undergo pose normalisation and dense mesh correspondence (another computationally heavy step) before statistical shape models are constructed using principal component-based methods. Regression models will then be applied to predict skeletal geometry from demographic variables such as sex, age, height, and body mass. This will allow sampling of representative full-body anatomies for specific demographic groups, e.g., 20-year-old female, 165 cm. The resulting models will provide population-representative skeletal geometries that preserve anatomical covariance across the body. These models will support the generation of representative occupants for next-generation HBMs and contribute to diverse, inclusive virtual safety assessment in crash simulations.