Functional limits of agreement applied as a novel method comparison tool for accuracy and precision of inertial measurement unit derived displacement of the distal limb in horses

Abstract

Over ground motion analysis in horses is limited by a small number of strides and restraints of the indoor gait laboratory. Inertial measurement units (IMUs) are transforming the knowledge of human motion and objective clinical assessment through the opportunity to obtain clinically relevant data under various conditions. When using IMUs on the limbs of horses to determine local position estimates, conditions with high dynamic range of both accelerations and rotational velocities prove particularly challenging. Here we apply traditional method agreement and suggest a novel method of functional data analysis to compare motion capture with IMUs placed over the fetlock joint in seven horses. We demonstrate acceptable accuracy and precision at less than or equal to 5% of the range of motion for detection of distal limb mounted cranio-caudal and vertical position. We do not recommend the use of the latero-medial position estimate of the distal metacarpus/metatarsus during walk where the average error is 10% and the maximum error 111% of the range. We also show that functional data analysis and functional limits of agreement are sensitive methods for comparison of cyclical data and could be applied to differentiate changes in gait for individuals across time and conditions.

Introduction

In humans, objective clinical gait analysis is widely applied (Wren et al., 2011) such as in cerebellar ataxia (Helwig et al., 2011, Serrao et al., 2012), spinal cord injury (Gil-Agudo et al., 2011, McDermott et al., 2010) and is considered crucial in the assessment of cerebral palsy (Chang et al., 2010).

Inertial measurement units (IMUs) provide an opportunity to collect stride series with the subject moving with fewer constraints than in a gait laboratory or on a treadmill. IMUs are increasingly applied in ambulatory monitoring of spatiotemporal kinematics in humans (Fong and Chan, 2010, Rueterbories et al., 2010, Saber-Sheikh et al., 2010, Schwesig et al., 2011) and for lameness assessment in horses (Barrey and Desbrosse, 1996, Keegan et al., 2011, Pfau et al., 2007, Thomsen et al., 2010).

In order to justify the use of limb-mounted IMUs to determine displacement (local position estimate) during gait analysis in horses, it is crucial to determine accuracy and precision. Accuracy and precision for IMU derived displacement have previously been determined for upper body landmarks in horses (Pfau et al., 2005, Warner et al., 2010) and centre of mass (CoM) in humans (Esser et al., 2009) as well as foot movement in humans (Mariani et al., 2010). However, the high impact, vibrations and dynamic range of rotation observed for the distal limb in the horse (Gustas et al., 2001, Wilson et al., 2003, Witte et al., 2004) constitute a particularly challenging environment for IMU sensor fusion approaches. A recent attempt at obtaining accuracy and precision from hoof mounted IMUs revealed measurement errors above 10% for latero-medial and dorso-ventral position estimates during walk and trot (Moorman et al., 2012).

While consensus amongst previous IMU validation studies (Mariani et al., 2010, Pfau et al., 2005, Schepers et al., 2010, Warner et al., 2010) provide root mean square (RMS) differences or accuracy defined as bias and limits of agreement as described by Bland and Altman (1986) and precision across all measurements throughout a stride, there is currently no agreement on a standardized statistical method to assess differences throughout a step cycle between 3D motion capture methods.

The aim of this study is to investigate the suitability of IMUs to quantify displacement of the distal limb in horses outside the gait laboratory. The objective is to compare displacement derived from distal limb mounted IMUs to motion capture (as the reference standard) using functional limits of agreement.

In contrast to Olsen et al. (2012) where we investigated accuracy and precision of temporal gait events extracted from limb mounted IMUs, here we compare spatial parameters between motion capture and IMUs.