A comparison of hip joint centre localisation techniques with 3-DUS for clinical gait analysis in children with cerebral palsy

Abstract

Functional calibration techniques have been proposed as an alternative to regression equations for estimating the position of the hip within the pelvic co-ordinate system for clinical gait analysis. So far validation of such techniques has focussed on healthy adults. This study evaluated a range of techniques based on regression equations or functional calibration procedures techniques in 46 children representative of those attending a major clinical gait analysis service against previously validated 3-D ultrasound techniques for determining the hip joint centre. Best agreement with ultrasound for the position of the hip within the pelvic coordinate system was found for the Harrington equations (mean 14 mm, sd 8 mm). Sphere fitting (mean ≈ 22 mm, sd 11 mm) performed better than transformational techniques applied locally (mean ≈ 33 mm, sd 12 mm) or globally (mean = 30 mm, sd 14 mm). The participants with cerebral palsy showed reduced range of movement compared with healthy adults. Differences between these results and studies modelling the effects of simulated noise on functional techniques can probably be attributed to differences between that noise and the soft tissue displacements that are actually occurring.

Introduction

Most clinical gait analysis still uses the regression equations suggested by Davis et al. [1] to estimate the position of the hip joint centre in the pelvic coordinate system. Alternatives are Harrington's regression equations [2] or three categories of functional calibration methods [3], [4], [5], [6], [7], [8], [9]. The earliest [10], [11] used sphere fitting algorithms based on the assumption that the co-ordinates of a thigh marker will map out a part of the surface of a sphere during movement if measured in the pelvic co-ordinate system. More recently transformational techniques in which joint parameters are optimised to give a best least squares fit between modelled and measured markers have been proposed [4], [7]. Global calibration applies transformational techniques to all the modelled joints simultaneously [12], [13]. These techniques have been used in several studies using simulated data [3], [4], [8], [9] and to establish repeatability [9], [13].

Studies comparing results with medical imaging data allow an assessment of the accuracy of the techniques. Two studies have compared results with bi-planar radiography [6], [11] and a further two with planar ultrasound [6] and three dimensional reconstructions from planar ultrasound [14]. The equations of Davis et al. [1] appear least accurate in predicting the hip joint centre position. Three recent studies suggest that functional methods are better [5], [6], [14] but the most recent suggests that Harrington's regression equations perform similarly well.

All of these studies have been on healthy adults. The aim of this study is to investigate the accuracy of both the regression equations and a range of functional calibration techniques on a representative sample of children attending for routine clinical gait analysis appointments most of whom have cerebral palsy.