Classification performance of sEMG and kinematic parameters for distinguishing between non-lame and induced lameness conditions in horses

Lindsay B St George; Tijn J P Spoormakers; Sarah Jane Hobbs; Hilary M Clayton; Serge H Roy; Jim Richards; Filipe M Serra Bragança

doi:10.3389/fvets.2024.1358986

Classification performance of sEMG and kinematic parameters for distinguishing between non-lame and induced lameness conditions in horses

Front Vet Sci. 2024 Apr 2:11:1358986. doi: 10.3389/fvets.2024.1358986. eCollection 2024.

Authors

Lindsay B St George¹, Tijn J P Spoormakers², Sarah Jane Hobbs¹, Hilary M Clayton³, Serge H Roy⁴, Jim Richards⁵, Filipe M Serra Bragança²

Affiliations

¹ Research Centre for Applied Sport, Physical Activity and Performance, University of Central Lancashire, Preston, United Kingdom.
² Department of Clinical Sciences, Equine Department, Faculty of Veterinary Medicine, Utrecht University, Utrecht, Netherlands.
³ Department of Large Animal Clinical Sciences, College of Veterinary Medicine, Michigan State University, East Lansing, MI, United States.
⁴ Delsys/Altec Inc., Natick, MA, United States.
⁵ Allied Health Research Unit, University of Central Lancashire, Preston, United Kingdom.

Abstract

Despite its proven research applications, it remains unknown whether surface electromyography (sEMG) can be used clinically to discriminate non-lame from lame conditions in horses. This study compared the classification performance of sEMG absolute value (sEMGabs) and asymmetry (sEMGasym) parameters, alongside validated kinematic upper-body asymmetry parameters, for distinguishing non-lame from induced fore- (iFL) and hindlimb (iHL) lameness. Bilateral sEMG and 3D-kinematic data were collected from clinically non-lame horses (n = 8) during in-hand trot. iFL and iHL (2-3/5 AAEP) were induced on separate days using a modified horseshoe, with baseline data initially collected each day. sEMG signals were DC-offset removed, high-pass filtered (40 Hz), and full-wave rectified. Normalized, average rectified value (ARV) was calculated for each muscle and stride (sEMGabs), with the difference between right and left-side ARV representing sEMGasym. Asymmetry parameters (MinDiff, MaxDiff, Hip Hike) were calculated from poll, withers, and pelvis vertical displacement. Receiver-operating-characteristic (ROC) and area under the curve (AUC) analysis determined the accuracy of each parameter for distinguishing baseline from iFL or iHL. Both sEMG parameters performed better for detecting iHL (0.97 ≥ AUC ≥ 0.48) compared to iFL (0.77 ≥ AUC ≥ 0.49). sEMGabs performed better (0.97 ≥ AUC ≥ 0.49) than sEMGasym (0.76 ≥ AUC ≥ 0.48) for detecting both iFL and iHL. Like previous studies, MinDiff Poll and Pelvis asymmetry parameters (MinDiff, MaxDiff, Hip Hike) demonstrated excellent discrimination for iFL and iHL, respectively (AUC > 0.95). Findings support future development of multivariate lameness-detection approaches that combine kinematics and sEMG. This may provide a more comprehensive approach to diagnosis, treatment, and monitoring of equine lameness, by measuring the underlying functional cause(s) at a neuromuscular level.

Keywords: ROC analysis; equine; gait analysis; kinematics; movement asymmetry; sensitivity; specificity; surface electromyography.

Grants and funding

The author(s) declare that financial support was received for the research, authorship, and/or publication of this article. This study was made possible with support from Morris Animal Foundation (Grant ID: D21EQ-406) and the British Society of Animal Science (BSAS) 2018 Steve Bishop Early Career Award. Open Access publication fees were funded by Utrecht University.