Hoof and leg ailments resulting in lameness are the third most costly problem of dairy producers in the United States costing over $500 million annually with over 15% affected cows. In addition, lameness is a critical animal well-being issue. The inability to routinely and objectively diagnose lameness has hindered development of effective management practices. A novel device has recently been developed that allows detection of lameness at early stages, often before there are visual signs of impairment. This device is called Lameness Detection System (LDS) and it measures the vertical forces of individual limbs as animals walk freely across the system. The goal of this proposal is to determine the effectiveness of the LDS in detecting lame dairy cattle in commercial dairy environments. The reliability of the software and the hardware systems will be assessed and the effectiveness of the lameness models that are based on the data from a university farm will be evaluated. If the previously developed lameness models turn out to be ineffective in commercial settings, attempts will be made to explain the observed deficiencies. This will determine the scope of a phase II project, in which the effects of farm characteristics on lameness models will be studied. It is anticipated that the results of the study will demonstrate that the Lameness Detection System (LDS) can reliably detect the onset of lameness in commercial dairies. OBJECTIVES: The specific objective of the proposed work is to measure the effectiveness of the lameness detection system (LDS) in detecting lame dairy cattle in three different commercial dairy farm environments. The accuracy of the automated lameness diagnoses from the LDS will be compared to diagnoses of an experienced veterinarian. The ultimate goal is to develop a commercial system that will reliably and accurately detect lameness at very early stages. This information will allow lameness to be treated more effectively, which, in turn, should reduce economic losses on dairy farms and improve the well being of dairy cows. Software has been developed to collect the data as cows walk freely over the LDS in a single file at the rate determined by animals without using mechanical gates. The reliability of the software to collect data and perform real time analysis to calculate lameness scores will be assessed. Software has been developed to handle transition problems where one cow is leaving, a second cow is entering and, for a short time, both cows are on the LDS. The software allows readings from the two cows to be separated into discrete readings for each animal. The reliability of this software will be assessed and the limitations will be identified. The reliability of the controller in the farm environment will be assessed. The LDS consists of two stainless steel platforms. The mechanical strength and characteristics of the platforms will also be assessed. Lameness models developed with data from a university farm will be evaluated for application to different commercial farms. To test applicability of models across farms, the same model will be applied to data from all three farms. By constructing a single database consisting of data from all three farms and modeling them simultaneously, it will be tested if there is a significant difference between farms in the ability of limb movement variables (LMV) to predict lameness scores. If the previously developed lameness models turn out to be ineffective in commercial dairy settings, the new lameness models will be derived that utilizes the data collected from the commercial dairy farms. It is anticipated that the phase I data from the three management practices implemented by the selected dairies will generate the required preliminary results concerning robustness of the LDS system across commercial units. This will lead to the design of other experiments (Phase II) that relate lameness prevalence, veterinary intervention, treatment and prevention to economic impact under three different management practices. The LDS will also facilitate research by allowing the efficacy of veterinary treatments to be measured objectively, and the time course of recovery to be accurately monitored. APPROACH: Lameness detection systems will be installed at three commercial dairies. The units will be placed in return alleys from the milking parlor. Measurements will be made for all animals after all milkings for six-months. The animals will be allowed to walk freely over the LDS in a single file at the rate determined by the animals. The three farms selected for the study have three different management schemes and goals, and have three different flooring systems. All farms are free-stall, feed totally mixed rations, and have automated management records that are available to the study. Gait and lesion examinations will be conducted over a 90-day period, with 15 new animals examined each week at each farm for a total of 180 animals per farm over the trial period. Clinical examinations will be performed blindly, i.e. the examining veterinarian will have no knowledge of the LDS scores. Results from examinations will be tabulated using a standardized form, which includes lesion number, size (length & width), type and location either inter-digital or in one of six claw regions. All gait analyses will be preformed on cement free-stall floors by walking cows at least 30 yards with turning in tight circles. New animals used each week will be selected randomly except that for each group of 15 animals, a minimum of 3 animals will be classified by the LDS as sound (score 1), mildly lame (score 2), and lame (score 3). Animals scored 4 or 5 and cows routinely trimmed and treated according to established management practices will be excluded. Lameness models developed with data from a university farm will be applied to data from all three farms. By constructing a single database consisting of data from all three farms and modeling them simultaneously, we can test if there is a significant difference between farms in the ability of LMVs to predict lameness scores. This can be done in PROC LOGISTIC (or GENMOD) using an indicator variable to identify the farm associated with each data record. While the models and the associated coefficient estimates may be different, the achieved ROCs might still be the same in all the farms. Individual ROC curves for each farm will be obtained and will be compared with each other. A formal test of significance will be conducted to see if the area under the ROC curves is indeed the same in all farms. If these ROCs are indeed the same it will testify the fact that the LDS is equally effective in all the selected dairy farms. If differences do exist, attempts will be made to explain these differences and the question of modifying lameness models based on farm characteristics will be addressed. This will shed light on the scope of the study the PIs will have to address in Phase II, where the effects of farm characteristics on lameness models will be studied. To test the reliability of software, the lameness scores will be calculated from the raw data stored in the log files of the controller and will be compared with the lameness scores calculated by the real time software. The reliability of the controller hardware and the platforms will be assessed by physically evaluating them at the end of the study
