Pelvis and femur shape prediction using principal component analysis for body model on seat comfort assessment. Impact on the prediction of the used palpable anatomical landmarks as predictors.

Léo Savonnet,Xuguang Wang,Serge Van Sint Jan,Sonia Duprey,Costin Daniel Untaroiu

doi:10.1371/journal.pone.0221201

Léo Savonnet, Xuguang Wang + Show 3 more

Open Access

https://doi.org/10.1371/journal.pone.0221201

Copy DOI

Abstract

A personalized pelvis and femur shape is required to build a finite element buttock thigh model when experimentally investigating seating discomfort. The present study estimates the shape of pelvis and femur using a principal component analysis (PCA) based method with a limited number of palpable anatomical landmarks (ALs) as predictors. A leave-one-out experiment was designed using 38 pelvises and femurs from a same sample of adult specimens. As expected, prediction errors decrease with the number of ALs. Using the maximum number of easily palpable ALs (13 for pelvis and 4 for femur), average errors were 5.4 and 4.8 mm respectively for pelvis and femur. Better prediction was obtained when the shapes of pelvis and femur were predicted separately without merging the data of both bones. Results also show that the PCA based method is a good alternative to predict hip and lumbosacral joint centers with an average error of 5.0 and 9.2 mm respectively.

Highlights

Estimating bone shapes and locating joint centers from external palpable anatomical landmarks (ALs) are required for biomechanical analysis in many applications [1],[2]
The objective of the present study is to evaluate a principal component analysis (PCA) based method to estimate the shape of pelvis and femur from a limited number of palpable ALs, as the present work is mainly motived by developing personalized finite element buttock-thigh model for assessing seating comfort
The prediction error decreased with the number of ALs

Summary

Introduction

Estimating bone shapes and locating joint centers from external palpable anatomical landmarks (ALs) are required for biomechanical analysis in many applications [1],[2]. This is particular true when a finite element human body model is used to estimate the internal strain in soft tissue for assessing seating discomfort (see a recent review [3]). An alternative is the kriging [5], which deforms a surface by controlling a set of points to match their target positions as performed by [6].

Objectives

Methods

Results

Conclusion