The statistical shape model (SSM) has been established as a powerful tool for medical image analysis [
1‐
6]. The goal of constructing a statistical shape model is to obtain a mean shape and description of variation from a collection of samples [
7‐
10]. The methods employed strongly depend on the chosen shape representation, which can be landmarks and meshes, medial models, Fourier surfaces, spherical harmonics, deformable models, wavelets description, non-uniform rational B-Splines and others [
11,
12]. The choice of the shape representation influences further processing and calculation and in that context landmark-based point distribution models have become popular and commonly used methods. Statistical shape models are usually used for the task of segmentation [
13,
14], but they could also be considered for finite element (FE) modelling [
15,
16] and automatic detection of shape and feature correspondences [
17,
18]. The SSM based techniques of medical image analysis have been applied to segmentation of bones [
19‐
22] but only few studies considered statistical shape of the calcaneus, cuboid, navicular and talus that constitute the four largest tarsal bones [
23‐
26]. There have been many studies considering modelling of foot bones. For example, Camancho et al. [
27] generated an anatomically detailed, three-dimensional reconstruction of a human foot from computed tomography (CT) images. They proposed an accurate representation of bone and soft tissues of foot. The presented method became a base for further development of a FE model of the human foot that could be used in quantifying morphometric characteristics between different foot types [
28]. Also, Liu et al. [
29] described rigid model-based 3D segmentation of joints imaged using magnetic resonance (MR) and CT images in order to examine their kinematics. Of all tarsal bones, talus has received most of attention. Leardini et al. [
30] proposed a geometric two-dimensional model of the ankle joint, which allows examining ankle stability. The presented model showed the path of calcaneus, ligament orientations, instantaneous axis of rotation, and conjugate talus surface profile as observed in the experiments. In their following work, [
31] they aimed at developing a model of the intact human ankle complex. The goal was to design the total ankle replacement which would better reproduce the physiological function of the joint. Such a model was used for FE analysis of total ankle replacement during the stance phase of gait [
32]. Contrarily, cuboid, and navicular were only broadly considered [
33].
All of the works mentioned above did not employ the SSM analysis. Recently, a SSM for calcaneus has been described, where an accurate SSM of calcaneus was proposed [
24]. The aim of this work was to extend that methodology to the case of the other three tarsal bones, namely cuboid, navicular, and talus. Additionally, it was of interest to ascertain whether SSM parametric characterisation can be used for classifying the particular tarsal bones.