Slide 1

The purpose of this study is to correlate a wireless body sensor device to subjective evaluation by three experienced veterinarians. In order for the wireless device to be of use to an equine practitioner, values obtained from the system that describe lameness must have meaning relative to the severity of the lameness observed. Therefore, this study will attempt to correlate the sensor values to those of the American Association of Equine Practitioners scale of 0-5 which is the standard in characterizing severity of lameness

(0= no lameness, 5= most severe lameness).

Correlation Analysis Between Three Experienced Veterinarians and a

Wireless Body Sensor System Used for Lameness Detection in Horses

Eric Dent

Faculty Mentor: Dr. Kevin Keegan

Figure 1: Four telemeterized, small body sensors, transceiver, PCI card, and laptop computer equipped with Matlab software. This is the complete system used for data collection.

Figure 2: Horse fitted with the wireless body sensor system. Sensor 1 is attached via Velcro to the halter directly over the poll. Sensor 2 is attached via Velcro to the pelvis directly between the tubera sacrale. Sensors 3 and 4 are attached to the cranial portion of the hoof wall with Elasticon. The handler holds the transceiver (5).

Figure 3: Example of the evaluation sheet filled out by each veterinarian after performing their lameness exam on the horse. The average of these values between the three veterinarians will be the final value used in conjunction with the wireless value for the correlation analysis.

The University of Missouri Veterinary Research Scholars Program was supported by funds from Merck-Merial, Pfizer and the MU College of Veterinary Medicine

So far, the study has shown great variation between clinicians in the subjective evaluation. Subjective evaluation averages shown below (Figure 4) were calculated from the evaluation sheets. Averaged numbers from three 30 second trials of wireless data were then correlated to subjective data using Pearsons correlation coefficient. A correlation value of 0.0732 (poor correlation) was obtained (Figure 5).

Figure 4: Subjective evaluation and wireless data from six horses.

Figure 5: Scatter plot of data with trendline.

Poor correlation has been observed using data collected so far. Poor agreement among clinicians is a contributing factor. In addition the wireless system presently quantifies forelimb and hindlimb lameness to one side only. Bilateral lameness will be overestimated on one side and underestimated on the other. Stronger correlation may be observed by using more rigid guidelines for subjective evaluation and by incorporating a partitioning of lameness between sides into the wireless data analysis. Also, an increase in number of horses studied may impact the results significantly. Preliminary agreement analysis between mean subjective evaluation and wireless analysis using on a single-limb basis is promising (see table below).

INTRODUCTION

The wireless system being used for the study consists of 4 small telemeterized body sensors which communicate with a transceiver held by the handler. The transceiver communicates with a wireless-enabled PCI card in a laptop computer (Figure 1). Horses obtained from the normal lameness caseload at the Veterinary Medical Teaching Hospital are being used for the study. Each horse is fitted with the wireless body sensor system (Figure 2) and trotted in a straight line on asphalt for 3 trials of 30 seconds during which data is collected. In addition, 3 experienced veterinarians (considered to be the gold standard for this study) perform a basic lameness evaluation on the horse at their own discretion which may include: walking, trotting, longeing and flexion tests. Veterinarians will be blinded to each other’s evaluation as well as to the results of the wireless device. Upon completion of the lameness examination to each individual’s satisfaction, the three veterinarians indicate in which (if any) leg(s) the lameness is present, and what grade of lameness (on the AAEP scale) it is present in each leg(s) (Figure 3). Data from the two sources will be analyzed using Pearson’s correlation coefficient. To date, data has been collected on six horses.

MATERIALS AND METHODS

PRELIMINARY RESULTS

DISCUSSION

agree

disagree

agree

disagree

agree

disagree

agree

disagree

agree

disagree

agree

Right hind

Left hind

Right front

Left front

Limb involved?

Horse number