Statistical Shape Modeling of the First Human Rib: Modeling Using CT Data and Demographic Predictors

Abstract

The first human rib plays a critical role in thoracic biomechanics and is increasingly relevant for improving simulations in human body models used in vehicle and traffic safety research. This thesis aims to develop a statistical shape model (SSM) of the first rib that captures morphological variability in terms of sex, age, height, and body mass index (BMI), based on computed tomography (CT) data. A total of 50 anonymized CT scans, sourced from a trauma hospital through the University of Michigan, were used from an available dataset of 104 samples. The workflow included segmentation in 3D-Slicer, landmarking in ANSA, and cortical bone thickness mapping using Stradview. To enable statistical analysis, a template rib surface was morphed to match each segmented rib using the Infepy Python library, so that all surfaces shared the same number and arrangement of nodes. Rib thickness values were also transferred onto these common nodes to combine shape and cortical thickness data. Generalized Procrustes Analysis (GPA) and Principal Component Analysis (PCA) were performed to align the ribs and identify dominant modes of variation, respectively. Linear regression was employed to examine relationships between principal component scores and demographic variables, and to predict shape and thickness parameters, which were then used to reconstruct corresponding 3D meshes. Four principal components showed statistically significant correlations (p ≤ 0.05) with demographic variables (age, sex, height, and BMI) in regression analysis. These components primarily encoded variation in rib size, cortical thickness, and subtle shape changes. However, age, sex, height, and BMI alone could capture only 15% of the total variability, indicating that these parameters were insufficient to fully predict rib morphology. Despite these limitations, the resulting model represents a valuable step toward anatomically realistic rib shape modeling and offers potential for enhancing personalized simulations in safety engineering. Additionally, PCA proved useful beyond model building by enabling quantification and visualization of how rib morphology varies across individuals.