Investigation of Thresholds for Asymmetry Indices to Represent the Visual Assessment of Single Limb Lameness by Expert Veterinarians on Horses Trotting in a Straight Line

Abstract

Defining whether a gait asymmetry should be considered as lameness is challenging. Gait analysis systems now provide relatively accurate objective data, but their interpretation remains complex. Thresholds for discriminating between horses that are visually assessed as being lame or sound, as well as thresholds for locating the lame limb with precise sensitivity and specificity are essential for accurate interpretation of asymmetry measures. The goal of this study was to establish the thresholds of asymmetry indices having the best sensitivity and specificity to represent the visual single-limb lameness assessment made by expert veterinarians as part of their routine practice. Horses included in this study were evaluated for locomotor disorders at a clinic and equipped with the EQUISYM® system using inertial measurement unit (IMU) sensors. Visual evaluation by expert clinicians allocated horses into five groups: 49 sound, 62 left forelimb lame, 67 right forelimb lame, 23 left hindlimb lame, and 23 right hindlimb lame horses. 1/10 grade lame horses were excluded. Sensors placed on the head (_H), the withers (_W), and the pelvis (_P) provided vertical displacement. Relative difference of minimal (AI-min) and maximal (AI-max) altitudes, and of upward (AI-up) and downward (AI-down) amplitudes between right and left stance phases were calculated. Receiver operating characteristic (ROC) curves discriminating the sound horses from each lame limb group revealed the threshold of asymmetry indice associated with the best sensitivity and specificity. AI-up_W had the best ability to discriminate forelimb lame horses from sound horses with thresholds (left: −7%; right: +10%) whose sensitivity was greater than 84% and specificity greater than 88%. AI-up_P and AI-max_P discriminated hindlimb lame horses from sound horses with thresholds (left: −7%; right: +18% and left: −10%; right: +6%) whose sensitivity was greater than 78%, and specificity greater than 82%. Identified thresholds will enable the interpretation of quantitative data from lameness quantification systems. This study is mainly limited by the number of included horses and deserves further investigation with additional data, and similar studies on circles are warranted.

Comparing Inertial Measurement Units to Markerless Video Analysis for Movement Symmetry in Quarter Horses

“…These "un-normalized" values are between 1.125 and 1.5 higher than the 8 mm threshold [25,26] and between 1.25 and 1.75 higher than the 4 mm threshold [25,26]. Compared to recently published thresholds [12], the LoA values presented here are of similar magnitudes.…”

“…based studies [7,22,23] and compared to thresholds derived from "non-lame" horses [12]. This is very encouraging for conducting larger-scale, multi-centre studies with the videoand IMU-based approaches compared here including straight-line, in-hand trot as well as circular lunge exercise, for which similar limits of agreement were found.…”

“…With an increasing diversity of camera-and IMU-based systems available [5,[9][10][11][12], the question of "comparability" is of growing importance. In particular, systems with minimal infrastructure requirements are opening the door for larger-scale and/or longitudinal followup studies.…”

“…In particular, systems with minimal infrastructure requirements are opening the door for larger-scale and/or longitudinal followup studies. Differences between the different systems in sampling rates [13], processing and filtering mechanisms [3,9,12,14,15], as well as stride segmentation [9,[16][17][18][19], introduce challenges for direct comparison. Yet, comparison without exact "stride matching" gives a more realistic assessment of how well different systems agree, since the stride selection is typically performed (semi-) automatically with limited possibility for intervention by the user.…”

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

“…For iHL, MinDiff and MaxDiff Pelvis and MinDiff Withers had comparable AUCs (>0.90) to the Hip Hike parameter, which indicated excellent discriminative power for iHL. Two recent studies have employed ROC analysis to evaluate the accuracy and associated threshold values of kinematic asymmetry parameters, measured using IMUs ( 33 , 40 ), for differentiating between non-lame and lame groups of horses, as defined using visual lameness evaluation by expert veterinarians ( 33 , 40 ). Our general findings agree with these studies, in that pelvic asymmetry parameters had a higher discriminative power (as quantified by higher AUC values) for detecting iHL than head and withers asymmetry parameters for detecting iFL ( 33 , 40 ).…”