Full length articleAn anatomically unbiased foot template for inter-subject plantar pressure evaluation
Introduction
Statistical analyses of pedobarographic images have traditionally consisted of discrete regional scalar extraction [1], [2], [3]. It has also been shown that, following spatial registration [4], these images can alternatively be analyzed in a continuous manner using Statistical Prametric Mapping (SPM) [5], [6]. SPM takes advantage of spatial correlation to conduct statistical inference across the entire pressure field. One advantage of this approach is that anatomical meaning is implicit in the resulting statistical images, so one avoids potential problems associated with explicit labeling of broad discrete regions [6].
Although registration generally ensures homologous data analysis, pedobarographic SPM analyses have, to date, used arbitrary subjects as registration templates. This is not ideal because subject peculiarities could conceivably introduce registration bias which could, in turn, affect the ultimate statistical results. The solution is to use a hypothetically average subject which, by definition of central tendency, constitutes an unbiased approach.
The main purpose of the current study was to construct a pedobarographic template that: (i) embodied healthy average morphology and pressure distribution, and (ii) was minimally biased by subject peculiarities. A secondary purpose was to demonstrate how this template could be used for population-based scientific studies.
Section snippets
Dataset
Pedobarographic data were collected from N = 104 healthy adult volunteers at the University of Münster (Table 1). Subjects performed five trials of self-paced walking for both the left and right feet. Data were recorded at 50 Hz using an EMED ST4 platform (resolution: 4 sensors/cm2) (Novel GmbH, Munich, Germany).
Preprocessing
Spatially maximal (‘peak’) pressure images were extracted and were then registered within-subjects and within-feet using mutual information maximization and a three-parameter (one-rotation,
Results
The final left and right linear templates were nearly identical (Fig. 2a), as were the nonlinear templates (Fig. 2b); both left and right feet exhibited regions of high pressure at the heel, central metatarsals, and hallux (Table 2). ANOVA found significant effects of METHOD (linear/nonlinear; p = 0.015) and REGION (heel, etc.; p < 0.001) but not of FOOT (left/right; p = 0.839). All other foot regions were associated with comparatively lower pressures. Since the left and right foot images were
Discussion
The present results demonstrate, firstly, that it is feasible to compute morphologically and functionally average pedobarographic templates for specific populations using image registration [4], [8] and average morphology techniques [9]. The resulting templates are minimally biased because, by the definition of central tendency, a population-wide average warp constitutes an unbiased estimate of the true deformation required to warp a given subject's foot to the population mean.
Present results
Acknowledgements
The authors thank Dirk De Clercq for his contributions and support of this study. Funding for this work was provided by Special Coordination Funds from MEXT, Japan and the Deutsche Forschungsgemeinschaft (Grant# RO2146/3-4).
References (15)
- et al.
Plantar pressure distribution measurements: technical background and clinical measurements
Foot Ankle Surg
(1997) - et al.
Image registration methods: a survey
Image Vision Comput
(2003) - et al.
Pedobarographic statistical parametric mapping: a pixel-level approach to foot pressure image analysis
J Biomech
(2008) - et al.
Nonlinear spatial warping for between-subjects pedobarographic image registration
Gait Posture
(2009) - et al.
Symmetry and limb dominance in able-bodied gait: a review
Gait Posture
(2000) - et al.
Plantar pressure distribution in normal, hallux valgus and hallux limitus feet
The Foot
(1999) - et al.
Pressure distribution underneath the human foot
Cited by (20)
Pedobarographic Statistical Parametric Mapping of plantar pressure data in new and confident walking infants: A preliminary analysis
2021, Journal of BiomechanicsCitation Excerpt :This was performed to reduce the impact of differences in shape and dimensions of the point clouds (Fig. 1), and to enhance further steps of data processing. To perform between subjects (BS) registration, an unbiased BS template was chosen as the foot with length and width closest to the mean length and width of all the feet analyzed (Pataky et al., 2011). The non-rigid coherent point drift (CPD) algorithm for point cloud registration was performed through a built-in MATLAB function, to allow the shape of the point clouds to change using a displacement field as transformation.
A state-of-the-art review of foot pressure
2020, Foot and Ankle SurgeryCitation Excerpt :Meanwhile, the in-shoe system benefits studying subjects’ motions on different shoes (e.g. heel size, shoes material, and shapes) to test on different footwear by experimental needs. The weakness is that for the in-shoe system the number of sensors installed is limited within the shoes’ coverage areas from as low as 3 [35,36] and up to 15 [10] placements. Both systems: plantar platform and the in-shoe have their own advantages and disadvantages.
An assessment of the information lost when applying data reduction techniques to dynamic plantar pressure measurements
2019, Journal of BiomechanicsCitation Excerpt :In order to perform the subsequent analyses, an anatomical correspondence is required between plantar pressures from different subjects. To obtain this correspondence for plantar pressure images, we employed the between-subject registration technique of Pataky et al. (2011). For the full plantar pressure measurement, we obtained correspondence using the between-subject registration and dynamic time warping algorithms from STAPP (Booth et al., 2018).