How to judge a tricky problem with objective tools.
As a young veterinarian with an interest in equine lameness, I looked up to, sought out the advice and direction of, and worked for equine practitioners seasoned in lameness detection and evaluation. Over the years, I learned many things, some from my colleagues and some just by plugging away, looking at as many lame horses as possible and following up with them as well as I could.
Figure 1: Stankey, a 5-year-old Quarter Horse gelding, being evaluated for lameness with the Lameness Locator by Dr. Brittany Hager, house officer intern at the University of Missouri College of Veterinary Medicine's Veterinary Medical Teaching Hospital. The system consists of 1) a head accelerometer sensor attached to the felt head bumper, 2) a right forelimb gyroscopic sensor in the pastern pouch, and 3) a pelvic acceleration sensor attached between the tubera sacrale with tape. Data are collected live as the horse is trotted in hand or ridden past the examiner.
Along the way, a few things became apparent to me. First, it was rare that two experienced practitioners observed a horse in motion the same way, and they often had difficulty expressing exactly what they were observing. Second, disagreement on selection of the affected limb between my mentors was frequent enough to be confusing, at least to me.
Given the dedication and experience of my mentors, I knew the reasons for these observations were not likely due to deficient skills. Many veteran lameness veterinarians have since confided to me that they have made the same observations. They all reported having their share of confusing cases and believe there must be a better way.
Lameness detection in horses is difficult to master, and, until recently, the lack of ability to objectively measure lameness's effect on movement prevented standardization. Thus, when veterinarians were taught lameness detection in veterinary school or by mentors in their first years of practice, they were likely taught one of a myriad of different ways proposed either in text or tradition.
The human eye also has its physical limitations. Small changes in the movement of a body part due to lameness may be missed or misinterpreted by the limited spatial and temporal resolution of the human eye. And, as humans, we subjectively express how the lameness appears to us and can be biased. This is the art of lameness detection. But our art is more properly directed to unraveling the complexities and interconnectedness of anatomy, pathology, imaging, treatment and prognosis—in other words, the skills and knowledge that separate equine practitioners from the crowd.
A search for a method of objective measurement began at the University of Missouri in 1993. We chose to study motion analysis with cameras and the treadmill. A wealth of information on the technique was already available from pioneers in equine motion analysis.
Figure 2: A Lameness Locator report with graphical display (top of report) and calculated lameness variables (bottom of report). The arrows on the left point to the forelimb evaluation. The arrows on right point to the hindlimb evaluation.
However, distilling the published research down did not point to a reliable set of measures. An unreasonably large number of possible measures needed to be studied: stride length, stride timing, limb swing patterns, head and hip motion, and so on. What exactly were the best motion parameters to measure to detect lameness?
We marked horses at every joint and prominent body part possible and collected motion data with high-speed cameras from large numbers of horses that were sound or lame, either naturally or from one of several different experimental models of lameness induction. We determined through analysis of these large data sets that vertical movement of the torso was the best approach for detecting lameness in horses. It was reassuring that others had suggested the same from earlier studies. But this was the first step.
Next, we had to determine what it is about this vertical motion that is characteristic of lameness. Study of the hundreds of motion signals collected over the years and consultation with engineers at the University of Missouri suggested that this was a fault detection problem, or, more specifically to engineers, a vibrational analysis for fault detection. When a horse is sound, the pattern of the vertical movement of its torso (at a trot) is a simple, symmetrical pattern. When a horse is lame, the vertical movement of its torso becomes more complex and asymmetrical with multiple possible patterns. This complexity and asymmetry of movement could be quantified.
In the year 2000, this approach to equine lameness evaluation was not entirely new, but it was impractical. Completing evaluations of horses trotting on treadmills, marked with reflective markers and filmed with high-speed cameras, was tedious and difficult work. No one was going to take the time and energy necessary using this method to evaluate a single horse. So, what good was it for a practitioner, where the ultimate need resides? Fault analysis of the vertical motion of the torso to detect lameness in horses was a curiosity, of interest at bioengineering or equine locomotion conferences but not to those on the front lines of equine lameness — equine practitioners.
However, further study suggested that the different patterns of vertical torso motion seen with various forms of lameness might mean something. In other words, there was untapped relevant information in the pattern of vertical torso motion. Different lameness foci resulted in different vertical torso movement patterns. Lameness with peak pain during impact or the first part of the stance phase of the stride, when the horse is coming down on the limb, may result in a vertical torso movement pattern different from lameness with peak pain during pushoff or the second part of the stance phase of the stride, when the horse is thrusting itself upward and forward by extending the limb. Also, compensatory patterns of head movement in hindlimb lameness and of pelvic movement in forelimb lameness contained additional information helpful to isolating lameness in horses.
Since 2000, the University of Missouri, in collaboration with engineers in Japan at the Hiroshima Institute of Technology, has been concentrating on developing this concept into a tool equine practitioners can use. The core belief was that this approach could be adapted to the field by using body-mounted inertial sensors and wireless transmission. First attempts were crude and impractical, employing large sensors with wires running to cell-phone-like devices strapped to a horse's back. Besides the incompatibility of horses and wires, the range was too short to be practical. Errors in translating the inertial sensor signals into forms compatible for the fault detection approach were too large. Transmission interference was significant. Much more work was needed.
Our quest to provide a clinically useful tool is now complete. Equine veterinarians and engineers at the University of Missouri have reached the point where they feel that computer-assisted lameness detection is a real possibility for practicing equine veterinarians. Years of development and validations have resulted in a body-mounted inertial sensor system specifically designed as a diagnostic aid to practicing equine veterinarians.
This tool involves noninvasively instrumenting the horse with three small (28-g) inertial sensors—one on the head, one on the pelvis and one on the right forelimb—in a process that takes about two minutes. The horse is then trotted in hand or ridden while movement data are collected. Movement data are transmitted at 200 Hz in real time to a hand-held computer at ranges up to 150 meters (Figure 1). The data are then analyzed using the aforementioned fault detection approach. The veterinarian is given information on lameness severity, single or multiple limb involvement, timing of lameness (impact or pushoff), and compensatory movement patterns common with equine lameness, as well as on the reliability of the data collected (Figure 2).
Presently, 18 of these body-mounted inertial sensor systems are out in the field in the United States, Europe and Canada, being tested and now used clinically by practicing equine veterinarians to evaluate lameness (see Table 1). Many of the suggestions for improvement from these early adopters have been incorporated into what is now called the Lameness Locator™. Current users range from solo ambulatory practitioners evaluating a few lameness cases a week to large referral hospitals with a lameness-predominant caseload and institutional veterinary teaching hospitals. Current users of the Lameness Locator find that it is most helpful under the following circumstances:
1. Evaluating horses with mild lameness
2. Evaluating horses with apparent multiple limb lameness
3. Evaluating horses with apparent compensatory lameness
4. Quantifying the effectiveness of peripheral nerve and joint blocks
5. Confirming the incidental nature of equivocal imaging abnormalities
6. Developing a further diagnostic approach based on type of the lameness (impact, pushoff)
7. Documenting musculoskeletal soundness in cases of poor performance
8. Monitoring convalescence or improvement after treatment.
As with any other diagnostic test or piece of medical equipment, there is a slight learning curve for most effectively using the Lameness Locator; the system will not suddenly turn a novice into an expert lameness evaluator. There is no substitute for dedicated experience and for performing a thorough physical and limb examination. Concentrating on the computer without looking at the horse is a sure way to miss something or disappoint the client. The inventors of the Lameness Locator stress that it is a valuable aid for practicing equine veterinarians, not a substitute. Most will find it easy to incorporate into their normal lameness evaluations. The Lameness Locator is a medical device and will only be made available to equine veterinarians.
Editor's note: The Lameness Locator is licensed by the University of Missouri to Equinosis LLC, a faculty-startup company launched in 2008 with the mission to assist and equip equine practitioners by developing wireless sensor solutions to collect biological data useful for the diagnosis of equine disease. Equinosis LLC is an angel-investor and Missouri Innovation Center-backed company with headquarters in Columbia (Missouri Life Sciences Incubator at Monsanto Place) and St. Louis (The Incubation Factory), Mo., with support from the National Science Foundation's Small Business Technology Transfer Research (STTR) program. The first commercial version of the Lameness Locator will be on display at the 2009 American Association of Equine Practitioners meeting in December in Las Vegas, Nev. Further information can be found at www.equinosis.com.
Dr. Kevin G. Keegan is an associate professor, Department of Veterinary Medicine and Surgery, College of Veterinary Medicine, University of Missouri, Columbia, Mo.