Article Text

Download PDFPDF

Use of force sensors to detect and analyse lameness in dairy cows
  1. M. Kujala, DVM1,
  2. M. Pastell, PhD2 and
  3. T. Soveri, DVM, PhD1
  1. 1 Department of Clinical Production Animal Medicine
  2. 2 Department of Agrotechnology, University of Helsinki, Pohjoinen Pikatie 800, Saarentaus, Helsinki, 04920 Finland


Force sensors were used to detect lameness in dairy cows in two trials. In the first trial, leg weights were recorded during approximately 12,000 milkings with balances built into the floor of the milking robot. Cows that put less weight on one leg or kicked frequently during milking were checked first with a locomotion scoring system and then with a clinical inspection. A locomotion score of more than 2 was considered lame, and these cows' hooves were examined at hoof trimming to determine the cause and to identify any hoof lesions. In the second trial 315 locomotion scores were recorded and compared with force sensor data. The force sensors proved to be a good method for recognising lameness. Computer curves drawn from force sensor data helped to find differences between leg weights, thus indicating lameness and its duration. Sole ulcers and white line disease were identified more quickly by force sensors than by locomotion scoring, but joint problems were more easily detected by locomotion scoring.

View Full Text

Statistics from

IN modern dairy husbandry, lameness has detrimental effects on the economics and welfare of dairy herds. According to Green and others (2002), lameness decreased milk yield from four months before, to five months after, a cow was diagnosed as lame, and many studies have shown how lameness or different types of hoof lesion can affect ovarian activity, udder health, milk yield and culling (Collick and others 1989, Barkema and others 1994, Garbarino and others 2004, Hultgren and others 2004). When studying lameness, two problems present themselves: the subjectivity of detecting lameness, and its aetiology. Sprecher and others (1997) described a locomotion scoring system that is used worldwide to detect lameness. More recently, technological approaches have been developed to detect lameness more objectively (Rajkondawar and others 2002a, b, Tasch and Rajkondawar 2004, Flower and others 2005, Flower and Weary 2006).

Objective methods are required to meet the needs of modern husbandry. Farms are becoming larger and farmers have less time to look after the cows properly. Detecting lameness is also one of the most important aspects of developing and validating new housing systems and floors.

An estimated 90 per cent of lameness is due to hoof problems, and sole ulcers are generally considered the most important cause of lameness in cattle (Murray and others 1996), and they are typically painful. However, hoof lesions do not always cause lameness (Smits and others 1992, Logue and others 1994, Manske and others 2002).

In this study, force sensors under the milking robot were used as an objective way to detect lameness and to follow its development (Pastell and others 2006), and to determine how the results correlated with hoof lesions and locomotion scores.


The study was carried out at the Suitia research farm, University of Helsinki. The farm has a typical Finnish loose-housing system with hard slatted floors with scrapers, and rubber mats in the cubicles. There are two 2 m-wide manure alleys, one next to the feeding barrier and the other between the rows of cubicles.

The prevalence of hoof lesions on the farm was examined between June and December 2003, before the trials. All 72 cows were hoof-trimmed in June and again between September and November, and the heifers were trimmed before parturition (seven in spring and 17 in autumn).

A system for automatically measuring the weight distribution between the four limbs of a cow while it was in the milking robot (DeLaval) began to be developed in autumn 2003, and the system was ready to collect data in May 2004 (trial 1). The system consisted of four strain gauge balances connected to an amplifier and a computer. The weight on each leg was recorded automatically with dedicated software by TestPoint (Capitol Equipment). The measurements were analysed and the data stored on the computer. The variables analysed were the mean weight on each leg, the total weight, the standard deviation of all the weights and the number of kicks during milking for each leg. A kick was recorded when the weight of a leg decreased to less than 20 kg. An automatic taring system was also used to minimise the effect of drifting. The system set the readings of the sensors at zero three times a day during the washing of the robot, when the robot was known to be empty. The measured weights were then compared with values previously obtained from the cow to identify any leg problems and were used in the first alarm list. All the milkings were recorded with digital video cameras and stored on a computer for one week.

When a cow stood outside the balances, it resulted in erroneous data that could be differentiated from correct values; these erroneous values were removed from the data and the corrected data were analysed by matlab (The Mathworks) (Pastell and others 2006). Suitia research farm had two 50-cow milking robots and one of them was used in all the trials.

The data were used to detect lameness in two ways: first, by using an alarm list (insensitive) on the computer that was based on a cow constantly keeping less weight on one leg, and secondly, by using graphs drawn with matlab (Figs 1, 2).


Measurements from the force sensors for a cow that rapidly reduced the weight it put on its right hindlimb owing to interdigital phlegmon; after three weeks isolation in a pen it had recovered and the distribution of weight on its hindlimbs was more nearly equal and its frequency of kicking had decreased


Measurements from the force sensors for a cow with a joint problem in its left hindlimb at the end of October and sole ulcers on both legs at the end of November. Its hooves were trimmed on December 2

Cows that put less weight on one leg or kicked frequently during milking were checked first with a locomotion scoring system (Sprecher and others 1997) and then by a clinical inspection. Locomotion was scored as follows: 1 Normal, the cow stands and walks with a level back posture, its gait is normal; 2 Mildly lame, the cow stands with a level back posture, but has an arched back posture when walking, its gait remains normal; 3 Moderately lame, the cow has an arched back posture when standing and walking, its gait is affected and it takes shorter steps with one or more legs; 4 Lame, the cow has an arched back posture all the time while standing, its gait is best described as one deliberate step at a time, and it favours one or more limbs; 5 Severely lame, the cow additionally demonstrates an inability to bear weight on one or more of its limbs. If cows had a locomotion score of more than 2 they were recorded as lame and their hoof trimming was checked to determine the cause. All lesions and other potential causes of lameness were recorded. During the same period all the heifers were also hoof-trimmed and/or checked near to parturition, and two and four months after parturition.

Leg problems were identified in both hindlimbs. In order to handle all the leg problems together in statistical testing the ratio of the weights carried by the hindlimb bearing less weight and the hindlimb bearing more weight was calculated. The ratio was calculated for the lame and healthy cows and a two-tailed independent sample t test was applied to determine whether there was a significant difference between the ratios of the lame and healthy cows. The t test was chosen, because it could be assumed that the two datasets came from a normal distribution, according to the central limit theorem (Clewer and Scarisbrick 2001).

In trial 2 the 50 cows that were milked in the robot were locomotion scored and videotaped six times at two-week intervals and a seventh time two months later, during winter 2005 to 2006. Some older cows with previous hoof problems were moved to the trial robot. In this trial, the cows were locomotion scored in a corridor leading to an old unused milking parlour. At three of these visits, the cows were scored by two or three observers and by the cattle handler on the farm. All the cows with a locomotion score of more than 2 recorded by at least one person, that is, the cows that walked more slowly or whose gait was somehow different, were hoof-trimmed and any lesions were photographed. During five of the seven visits, some of the cows with a locomotion score of 1 to 2 were chosen randomly, so that at least 40 per cent of all the cows (about 20) were examined and hoof-trimmed. The correlation between the locomotion scores of 49 cows, recorded by three independent observers, was also calculated. The locomotion scores and curves were examined independently, and all suspected lame animals were hoof-trimmed and inspected. Data from the locomotion scores and hoof trimming were then compared with the weight graphs from the automatic weighing system.


In the spring before the trials began, the hoof lesions observed were 61 per cent haemorrhages, 17 per cent white line separation or disease, 3 per cent sole ulcers and 52 per cent heel horn erosion or interdigital dermatitis; in the autumn these proportions had changed slightly to 67 per cent haemorrhages, 15 per cent white line disease, 1 per cent sole ulcers and 37 per cent heel horn erosion or interdigital dermatitis.

The development of the weighing system began in autumn 2003, but data were collected only from May 2004 to the end of the year. Refinements to the system continued until the end of February 2005. Over 12,000 milkings were recorded during trial 1, and 15 lame cows were identified with 16 hoof problems; seven white line diseases, eight sole ulcers, and one tyloma with interdigital phlegmon. Two cows that became lame as a result of injuries were easily found by the weighing system. No digital dermatitis was observed.

The results from two of the lame cows are shown in Figs 1 and 2. The cow in Fig 1 rapidly reduced the weight that it put on its right hindlimb; the diagnosis was interdigital phlegmon. It was isolated in a sick pen until it had recovered after three weeks, when the weight it put on the affected leg and its frequency of kicking had returned to normal.

The cow in Fig 2 had had a joint problem from September 27, 2004 (when it was first inspected) until October; on December 2, a sole ulcer was found on both its hindlimbs. The recovery can be seen in the centre of the graph.

Fig 3 shows the distribution of the ratios of the weight placed on the cows hindlimbs by 18 lame cows and 44 healthy cows. The ratios for the lame cows were calculated when they were diseased or injured and the ratios for the healthy cows were calculated from their average leg weights in August 2004. Almost all the lame cows placed much more weight on one leg than the other, whereas the healthy cows were more evenly balanced, although some of them also put less weight on one leg. There was a significant (P<0·01) difference between the healthy cows and the lame cows.


Distribution of the ratio of the weights placed on their hindlimbs by 18 lame cows and 44 healthy cows, measured by the force sensors

The second trial was undertaken in winter 2005/06 by three different people. When there were obvious hoof lesions or visible injury or inflammation to the joints, the locomotion scoring system correlated well with the hoof lesions. In this trial, sole ulcers, white line disease and clear overgrowths were nearly always given a score of more than 2; one bilateral hoof problem was missed but it was detected in the graphs. Cows developed haemorrhages either with or without an association with their scores, but were usually scored 2. When using this subjective detection method, all the observers recognised lameness only with cows with scores of 3 or more.

In total 45 cows were scored seven times (315 scores), and 26 of them were taken for hoof trimming, 11 of them twice and 15 of them three or more times (104 scores). Of these cows, 24 (44 scorings) had scores of more than 2. The causes of the lameness in these cows were sole ulcers (38 per cent), serious white line problems (9 per cent) and abnormal claw shape (9 per cent); the other 44 per cent were cases in which it was difficult to determine the real cause. These cows had concurrent mild joint problems, haemorrhages, white line separation or no visible problem in the soles or limbs.

Fifty per cent of the cows with a score of more than 2 had a sole ulcer at least once, sometimes after severe white line disease or together with the disease. Three cows with sole ulcers (one mild, one bilateral and one unilateral) were given a score of 2. Correlations between the scores assessed by the three veterinary observers were 62 per cent, 66 per cent and 75 per cent.

According to the balances, the prevalence of sole ulcers in the cows was 47 per cent during the first trial and 43 per cent during the second trial. During the first six observations, according to 50 graphs, there were 16 clear and 12 suspected lame animals. Of the 16 clear cases, eight had sole ulcers, two had white line disease, five had abnormally shaped claws or injuries and one had no specific cause. Of the 12 suspected cases a definite cause was found in only two, and the causes of the lameness in the other 10 cases were as uncertain as their graphs.

Table 1 shows that the floor balances usually detected hoof problems more quickly. Hoof problems in one cow with severe white line disease in the left hoof and a small sole ulcer in the right hoof were completely missed by locomotion scoring. Of the sole ulcers, the floor balances detected all but one cow that had sole ulcers on both back legs. Of the four cows with severe white line problems, the weight graphs showed three clearly and the other was detected over a longer period of time. Joint problems were detected more satisfactorily and more quickly by locomotion scoring.


Causes of lameness

Lameness always affected the back legs, in agreement with earlier studies in which lesions have been found mainly in the hind hooves (Andersson and Lundström 1981, Murray and others 1996, Manske and others 2002). Sole ulcers have been considered the most common cause of lameness (Murray and others 1996, Manske and others 2002), and the prevalence of sole ulcers determined by the floor balances was 47 per cent among the lame cows during autumn 2004, and 43 per cent during winter 2005. In Sweden, Manske and others (2002) reported a prevalence of 28 per cent. The prevalence of 47 per cent in trial 1 was particularly high for heifers, possibly owing to the very wet summer. The cows and heifers were pastured throughout summer 2004 in wet and dirty conditions, the rainfall that year being the heaviest for 30 years. All but one heifer that calved during the summer developed a sole ulcer, and it developed a white line separation. The softening effects of a wet environment on the claws has been shown by Enevoldsen and others (1991) and Borderas and others (2004), who found that cows with softer claws tended to have more severe claw lesions.

The high prevalence of sole ulcers (43 per cent) in trial 2 can be explained by the fact that older cows with previous problems were moved to the trial robot in autumn 2005. The results from the floor balances correlated very well with sole ulcers, severe white line disease and bad injuries. It was more difficult for the machine to identify lameness in cows with bilateral sole ulcers.

Locomotion scoring

Cows should be scored where they can be seen walking an adequate distance and without anything unusual happening around them. In this trial the cows were used to being milked by a robot, and thus walked rather differently when using an unfamiliar narrow passage from the milking parlour. It was, however, the only place on the farm where the scorings could be done. The rather low correlations between the scores of the three observers (62 per cent, 66 per cent and 75 per cent) show how subjective the procedure can be. By using more people scoring simultaneously, the cows that had a problem could have been identified more accurately. Nevertheless, locomotion scoring is subjective and demands great experience to give reproducible results.

Weighing system

When first trying to identify the lame cows, false positive results were quite often obtained. However, high sensitivity is very important. Cows behave unpredictably; sometimes a lower weight recorded on one leg was caused by a cow standing with both its hindlimbs on the same weighing plate, a situation that could only be differentiated from it actually putting less weight on one leg by direct observation or by examining the videos.

After it had been fully developed, the force sensor system with graphs worked well in detecting lame cows and was also useful in following their recovery. Severely lame cows, with locomotion scores of 4 or 5 could easily be detected with the robot when the computer put them on the ‘alert list’ for examination at the next milking. However, the system could have been more sensitive. In random hoof trimming, lesions were found that probably made the cows uncomfortable or caused them mild pain, but the floor balances usually did not detect them until the lesions became more severe. The preventive effect of hoof trimming on lameness could be seen in the graphs. The balances were quick to detect problems in one leg, but were sometimes slower when the problem was bilateral. More experience of looking at the curves may help to develop an ‘eye’ for detecting lame cows more quickly. The data should also make it possible to develop a model for detecting cases of lameness.

The floor balances performed well in comparison with locomotion scoring. In trial 2, they found the same problems as the observers, but slightly more accurately, and in eight cases, one of which was severe, they detected the problem more quickly. From an economic perspective the early detection of problems is important and because they were detected more quickly, the floor balances probably shortened the recovery time of the cows by several days.

The balances were not as good as human observers in detecting mild joint problems. This is easy to understand, because the cows were standing still on the sensors, whereas walking involves the hock joint, the movement of which readily reveals the problems. With the cows that were very stable during milking, it took longer to detect the problems. The graphs also showed that some previously lame cows could walk normally during the locomotion test, but become badly lame shortly afterwards. It is difficult to say why this happened, but it does show that a problem can be evaluated more adequately by long-term measurements.

Measurements also need to be made over a sufficiently long period to observe the cows properly, because bilateral problems, in particular, are more difficult to identify. Cows are often affected bilaterally, and frequently escape detection until hoof trimming (Logue and others 1994). The cows walked more slowly when they had problems in both back legs, and sometimes the pain shifted from leg to leg, depending on the examination period. These severe bilateral problems can probably be detected more easily by eye than by using floor balances. On the other hand, severe bilateral hoof problems make cows lie down longer without walking to the robot, and the longer time between milkings will be recognised by the milking robot's computer.

Because of the absence of infectious hoof diseases in the second trial, it was not possible to determine whether it would have been easy to detect them. Infectious diseases such as digital dermatitis usually make cows limp, so the balances would probably have detected the problem.

The floor balances were a good objective method for detecting lameness and monitoring the recovery process. When compared with locomotion scorings they were usually better, and they detected severe hoof problems more quickly. The only drawback was the need for the balances to be checked regularly; they were under heavy mechanical stress and frequently needed to be repaired. Technological improvements are needed.


The authors thank the Suitia research farm, especially cow manager Daniel Björkman, who is greatly missed after his sudden death in 2006, and veterinary students Frans Lahdenranta and Sanni Kallio, who helped with the scorings. Financial support from the Interrobo and Kartek projects, the Walter Ehrström Foundation and Suomen Rehu is gratefully acknowledged.


View Abstract

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.