Georgios Oikonomou

A fully automated 2-dimensional imaging system that uses machine learning to produce real-time mobility scores has been developed and previously externally validated using human mobility scores and foot lesion records as ground truth. This randomized controlled trial evaluated the effect of integrating this system into an early detection and prompt treatment lameness management protocol on a large dairy farm in the UK A total of 419 multiparous cows ≤30 d-in-milk (DIM) were randomly allocated to either a control (CON) group (n = 208), managed under the farm's standard protocol or an intervention (AUTO) group (n = 211). The CON protocol consisted of routine trims at early (approximately 80 DIM) and mid-lactation (approximately 180 DIM), and examination of cows identified as lame by farm staff. In addition to the CON protocol, weekly automated scores were obtained for AUTO cows. Any AUTO cow exceeding the pre-defined threshold (≥50, on a scale of 0 to 100) or those with a ≥20 points increase in absolute scores during the last 2 weeks were scheduled for examination and treatment. Lameness scores from monthly human mobility scoring sessions were compared between groups using Fisher's exact tests or Chi-squared tests, with relative risks (RR) and odds ratios (OR) calculated. Trimming events, foot lesion prevalence and severity, and number of hoof block applications required were compared between groups using Poisson regressions and Chi-squared tests. The effect on weekly average milk yield was assessed with linear mixed effects models. Culling hazard was assessed using Cox proportional hazards regression (COXPHR). Time to 1st artificial insemination (AI) and time to conception by 150 DIM were assessed with COXPHR, whereas odds for pregnancy to the 1st AI were assessed with binary logistic regression. Cows in the AUTO group had a lower proportion of cows that developed severe lameness (2.0% vs. 7.9%, RR = 0.25; 95% CI: 0.09-0.66; OR = 0.24; 95% CI: 0.08-0.69) and chronic lameness (3.9% vs. 9.8%, RR = 0.40; 95% CI: 0.18-0.91; OR = 0.38; 95% CI: 0.16-0.88) compared with CON cows. Cows in the AUTO group underwent 2.67 trimming events per cow compared with 1.83 in the CON group during the study period (as estimated marginal means). At the 180 DIM routine trim, the AUTO group had a higher proportion of lesion-free cows (22.4% vs. 12.0%) and a lower proportion of cows with moderate lesions (16.0% vs. 25.3%). The small subset of second-parity cows in the AUTO group had higher odds of conception to 1st AI (OR = 7.6; 95% CI: 1.6-36.7) and a greater hazard of conception by 150 DIM (HR = 3.1; 95% CI: 1.3-7.3) compared with their CON counterparts. No differences were detected for weekly average milk yield or culling risk. Our findings indicate that automated mobility monitoring can improve lameness control programs by reducing severe and chronic lameness and improving mid-lactation foot health in cows.

The digital dermatitis index (DDI), a genetic index for digital dermatitis (DD) resistance has been previously associated with presence of DD in Holstein cows. The aim of this study was to further investigate this association in a population of Holstein heifers. Two cohorts of heifers from one farm were enrolled. For the first cohort, DD lesions from the hind feet were recorded with the animals standing, using at telescopic mirror once a month over a period of 3 mo. For the second cohort, DD lesions were recorded from the hind feet by picking them up in a foot-trimming chute once a month over a period of 4 to 6 mo. The M-stage scoring system was used, and 4 binary variables were created. The first binary variable created was grouping animals that never had any DD lesion (M0) and animals that had at least one lesion recorded throughout the study. Similar binary transformations were repeated for active (M1, M2, M4.1) lesions, chronic (M3, M4), and M2 lesions separately. Genomically estimated predicted transmitting abilities (gPTA) for the DDI were available for 519 heifers of the first cohort and 361 of the second. The DDI gPTA were standardized (rescaled to represent distance from the mean expressed in SD) and were also grouped into terciles (i.e., low-, medium- and high-genetic-merit groups) within each cohort. Univariable analysis was used to investigate the association between the 4 binary lesion-presence variables and the DDI gPTA for each cohort separately and multivariable logistic regression models were fitted for the second cohort. Survival (time to event) analysis was performed using the second-cohort dataset by fitting a parametric proportional hazard regression model using the Weibull distribution for interval censored data, with DDI and season at enrollment remaining as explanatory variables. There was no difference between low-, medium-, and high-genetic-merit heifers regarding the presence of any, active, or chronic lesions in the first cohort. However, for a one-unit increase in the DDI gPTA, the odds ratio (OR) of M2 lesion presence was 0.68 (95% CI: 0.50-0.92). Multivariable logistic regression models using data from the second cohort showed that for every one-unit increase in the DDI gPTA the OR for any, chronic, active, and M2 DD lesion presence were 0.64 (95% CI: 0.50-0.80), 0.60 (95% CI: 0.47-0.76), 0.55 (95% CI: 0.42-0.70), and 0.59 (95% CI: 0.41-0.84), respectively. The proportional hazard model showed that a one-unit increase in the DDI gPTA resulted in a hazard ratio of 0.70 (95% CI: 0.60-0.82). The model-adjusted median age at first DD event was 306 d (95% CI: 284-328) for heifers with a DDI of -1 compared with 377 d (95% CI: 348-411) for heifers with a DDI of +1. We show significant association between DDI gPTA and presence of DD lesions. Lower-genetic-merit animals developed a DD lesion at a younger age compared with their higher-merit counterparts. Genetic selection using the DDI could gradually reduce the disease prevalence during the rearing period of replacement animals.

Early detection and prompt treatment of lame cows are crucial for proactive lameness management. This study aimed to evaluate a fully automated 2-dimensional imaging system for real-time lameness detection using artificial intelligence. Data were collected from 11 dairy farms in the UK Four trained veterinarians performed 42 mobility scoring sessions using a 0-3 4-grade scoring system, with scores 2 and 3 representing lameness. On each session, individual weekly average scores were calculated. This resulted in 40,116 paired human mobility scores (HMS) and weekly average mobility scores generated using artificial intelligence (AIMS) matched to a cow ID. Categorical agreement for the 4-grade scale was estimated by calculating the weighted Cohen's kappa (κ_w) and Gwet's agreement coefficient (AC₂), and for the 2-grade scale (nonlame vs. lame) by calculating the percentage agreement (PA), unweighted Cohen's kappa (κ) and Gwet's coefficient (AC₁). A trained veterinarian recorded the presence and severity of any lesion of 2,515 cows, which also had an AIMS assigned. A subset of 758 cows were also assigned an HMS 1-3 d before trimming. Sensitivity (Se), specificity (Sp), and accuracy (Acc) were calculated to describe the system's and human's ability to detect cows with foot lesions. Additionally, automated mobility scores were retrieved for cows with foot lesion records up to 30 d before trimming. Linear mixed effects models (LMM) were built to assess the association of the lesion status at trimming with the daily scores. The average (mAVG), maximum (mMAX), minimum (mMIN) and the percentage of scores that a cow was identified as lame (mPLS) during the 30 d before foot trimming were calculated and their Se, Sp and Acc in detecting foot lesions were determined. Lastly, longitudinal data were obtained from 143 cows tracking daily scores from 5 to 64 DIM. The association of lesion status at the early lactation routine trim (ELRT) with the daily scores was assessed by fitting LMM. Regarding the 4-grade scale agreement between HMS and AIMS, κ_w (0.24-0.34) represented fair agreement, whereas AC₂ (0.81-0.93) almost perfect agreement. For the 2-grade scale agreement, PA was consistently above 80%, κ (0.23-0.38) represented fair agreement, and AC₁ (0.76-0.83) showed substantial to almost perfect agreement. The AIMS detected cows bearing severe lesions with Se = 0.53 and Sp = 0.74, whereas the HMS achieved Se = 0.60 and Sp = 0.78. Using optimal thresholds for mAVG, mMAX, mMIN, and mPLS, the system achieved higher Se than HMS. Moreover, cows with severe lesions had increased scores from 23 d before trimming compared with cows with mild and moderate lesions. Longitudinal data showed that cows with severe lesions at ELRT had higher mobility scores during the first 60 DIM compared with those with mild or moderate lesions. Overall, the system's performance was comparable to that of experienced human assessors in detecting lame cows and cows with foot lesions. Finally, its capability to detect mobility changes before the development of severe lesions highlights its potential for early intervention, which could enhance lameness management in dairy herds.

Background: Internal teat sealants are commonly used at drying off, with or without intramammary antibiotics, to reduce the risk of mastitis. Both full and partial cannula options are available on most teat sealant and antibiotic tubes, but little evidence exists to support the selection of one option over the other. Methods: A total of 287 Holstein cows from three UK farms were enrolled in the study. Cows were randomly allocated to receive either full or partial insertion of internal teat sealant at drying off (plus the same insertion type of intramammary antibiotic if required by farm protocols). Somatic cell count and clinical mastitis data were collected, along with lactation number and calving season, and analysed using multivariable regression modelling. Results: Insertion type was not associated with high postcalving somatic cell counts (SCCs; >200k cells/mL), new infection rates (low to high SCC change across the dry period), cure rates (high to low SCC change across the dry period) or clinical mastitis cases. Limitations: Only pedigree Holsteins were included; results in other breeds could differ due to teat size. Conclusion: Insertion type was not associated with high SCC or increased postcalving mastitis cases.

Key factors such as stage of lactation, parity, and body fat reserves have been associated with the digital cushion thickness (DCT), however, there are discrepancies between the results of previously published studies. The objective of this study was to examine the association of stage of lactation, body fat reserves, parity, and lesion incidence with DCT in a large cohort of intensively monitored cows. Across 4 UK farms, 2,352 cows were prospectively enrolled and assessed at 4 time points: before calving (T1-Precalving), immediately after calving (T2-Calving), in early lactation (T3-Early), and in late lactation (T4-Late). At each time point, BCS was recorded, the presence of sole lesions (sole ulcers and sole hemorrhage) and white line lesions was assessed by veterinarians, and an ultrasound image was taken to retrospectively measure the backfat thickness (BFT) in the pelvic region and the digital cushion on the hind left lateral claw. Mixed effects multivariable linear regression models, with the cow as a random effect, were fit to examine the association between the explanatory variables and DCT. The explanatory variables tested were farm, parity, stage of lactation, BCS, BFT, height, the presence of a lesion at the time of measurement, the chronicity of a lesion during early lactation, the predicted maximum daily milk yield, and the rate of milk production rise in early lactation. Stage of lactation and farm were both associated with DCT; however, an interaction was present, and this DCT pattern of change was farm-dependent. Two distinct patterns emerged; one indicated the nadir to occur shortly after calving, the other indicated the nadir to occur during early lactation. Neither BFT nor BCS were significantly associated with DCT. Heifers displayed thinner digital cushions compared with multiparous cows; however, this effect was dependent on the stage of lactation, with heifers having a thinner digital cushion up until late lactation, by which time DCT was commensurate with multiparous animals. Sole lesions and white line lesions at the time of measurement were associated with DCT (sole lesion: estimate = -0.07 mm, 95% CI = -0.14-0.00; white line lesion: estimate = 0.28 mm, 95% CI = 0.15-0.42).

Digital dermatitis (DD) is a polybacterial disease endemic to most UK dairy farms. It poses a major financial and welfare threat and is characterized by high incidence and recurrence rates. We aimed to investigate the association between the UK EBV for resistance to digital dermatitis, the digital dermatitis index (DDI), and the frequency of DD, heel horn erosion (HHE), and interdigital hyperplasia (IH) in a population of Holstein dairy cows. We enrolled and genotyped 2,352 cows from 4 farms in a prospective cohort study. Foot lesion records were recorded by veterinary surgeons for each animal at 4 time points during a production cycle, starting at approximately 2 mo before calving and ending in late lactation. Importantly, these records were not used in the calculation of the DDI. Lesion records were matched to the animal's own DDI (n = 2,101) and their sire's DDI (n = 1,812). Digital dermatitis index values in our study population ranged from -1.41 to +1.2 and were transformed to represent distance from the mean expressed in SD. The relationship between the DDI and the presence of DD was investigated using a logistic regression model, with farm, parity, and a farm-parity interaction fitted as covariates. A multivariable logistic regression model was fitted to evaluate the relationship between HHE and DDI with farm fitted as a covariate. Finally, a univariable logistic regression model with DDI as explanatory variable was used to investigate the relationship between IH and DDI. The odds ratio of an animal being affected by DD was 0.69 for 1 SD increase in the animal's DDI (95% CI = 0.63-0.76). The odds of HHE and IH were 0.69 (95% CI = 0.62-0.76) and 0.58 (95% CI = 0.49-0.68) respectively for 1 SD increase in DDI. The adjusted probability of DD was 32% (95% CI = 27-36%) for cows with mean DDI value of 0, while it was 24% (95% CI = 20-29%) in cows with a DDI value of +1. Sire DDI breeding values were standardized in the same way and then binned into terciles creating an ordinal variable representing bulls of high, medium, and low genetic merit for DD resistance. The daughters of low genetic merit bulls were at 2.05 (95% CI = 1.60-2.64), 1.96 (95% CI = 1.53-2.50), and 2.85 (95% CI = 1.64-5.16) times greater odds of being affected by DD, HHE, and IH, respectively, compared with the daughters of high genetic merit bulls. The results of this study highlight the potential of digital dermatitis genetic indexes to aid herd management of DD, and suggest that breeding for resistance to DD, alongside environmental and management control practices, could reduce the prevalence of the disease.

The aims of this study were to assess (1) the variation of protein metabolism biomarkers and factors affecting them during the transition period, (2) the association of each biomarker with skeletal muscle reserves and their changes, and (3) the association of these biomarkers with postpartum health, colostrum quality, reproduction, and milk production. For this purpose, 238 multiparous Holstein cows from 6 herds were used in a prospective cohort study. Plasma concentrations of 3-methylhistidine (3-MH) and 1-methylhistidine (1-MH) and serum concentrations of total protein (TP), albumin (ALB), urea nitrogen (BUN), and creatinine (SCR) were determined for each cow at -21, -7, 7, 21, and 28 d relative to calving. Clinical diseases were recorded during the first 28 d postcalving, and presence of subclinical ketosis (scKET) was investigated at 7 and 21 d. Colostrum quality was estimated by Brix refractometry. Reproduction data by 150 d in milk (DIM) and milk production records were also available. Linear mixed models including the fixed effects of time point, herd, parity, body condition score (-21 d), duration of dry period and postparturient diseases were fitted to assess the variation in each biomarker's concentration. The association between the biomarkers' concentration during the prepartum period with the odds for each postparturient disease and for a combined trait (CD_1-28), defined as the presence of at least one clinical condition during the first 28 d after calving, were assessed with separate binary logistic models for time points -21 d and -7 d. The relationship of each biomarker's concentration with longissimus dorsi thickness (LDT) and the changes in LDT (ΔLDT) was assessed with pairwise correlations. Separate general linear models were used to assess the association of each biomarker with colostrum Brix values and milk production traits. Finally, the associated hazard for first artificial insemination (AI) and for pregnancy by 150 DIM (PREG_150DIM) was assessed with Cox proportional hazard models, whereas odds for pregnancy to the first AI (PREG_1stAI) were assessed with binary logistic models. The level of 3-MH was affected mainly by herd, time points, and their interaction. Higher 3-MH was associated with increased odds for metritis and CD_1-28, increased hazard for PREG_150 DIM and with increased milk production. 1-Methylhistidine was affected mainly by herd, scKET and occurrence of displaced abomasum. Higher 1-MH was associated with better colostrum quality, increased odds for scKET, increased hazard for first AI by 150 DIM and with decreased milk production. Both 3-MH and 1-MH were weakly to moderately negatively correlated with LDT and moderately to strongly negatively correlated with ΔLDT at the corresponding time periods. Additionally, higher TP was associated with increased odds for metritis and CD_1-28 and increased milk production, while higher ALB was associated with increased odds for scKET and increased milk production. Moreover, higher BUN was associated with decreased odds for scKET, increased odds for PREG_1stAI and increased milk production. Higher SCR was associated with decreased odds for retained fetal membranes, metritis, and CD_1-28. Periparturient protein metabolism is significantly associated with postpartum health, colostrum quality, reproduction, and milk production; mechanisms involved require further investigation.

Monitoring body condition score (BCS) is a useful management tool to estimate the energy reserves of an individual cow or a group of cows. The aim of this study was to develop and evaluate the performance of a fully-automated 2D imaging system using a machine learning algorithm to generate real-time BCS for dairy cows. Two separate data sets were used for training and testing. The training data set included 34,150 manual BCS (MAN_BCS) assigned by 5 experienced veterinarians during 35 visits in 7 dairy farms. Ordinal regression methods and deep learning architecture were used when developing the algorithm. Subsequently, the testing data set was used to evaluate the developed BCS prediction algorithm on 4 of the participating farms. An experienced human assessor (HA1) visited these farms and performed 8 whole-milking-herd BCS sessions. Each farm was visited twice allowing for 30 d (±2 d) to pass between visits. The MAN_BCS assigned by HA1 were considered the ground truth data. At the end of the validation study, MAN_BCS were merged with the stored automated BCS (AI_BCS) resulting in a testing data set of 9,657 single BCS. A total of 3,817 cows in the testing data set were scored twice 30 d (±2 d) apart and the change in their BCS (ΔBCS) was calculated. A subset of cows in one farm were scored twice on consecutive days to evaluate the within-observer agreement of both the human assessor and the system. The manual BCS of 2 more assessors (HA2 and HA3) were used to assess the inter-observer agreement between humans. Finally, we also collected ultrasound measurements of backfat thickness (BFT) from 111 randomly selected cows with available MAN_BCS and AI_BCS. Using the testing data set, intra- and inter-observer agreement for single BCS and ΔBCS were estimated by calculating the simple percentage agreement (PA) at 3 error levels, and the weighted kappa (κ_w) for the exact agreement. A Bland-Altman plot was constructed to visualize the systematic and proportional bias. The association between MAN_BCS and AI_BCS and the BFT was assessed with Passing-Bablock regressions. The system had an almost perfect repeatability with a κ_w of 0.99. The agreement between MAN_BCS and AI_BCS was substantial, with an overall κ_w = 0.69. The overall PA at the exact, ± 0.25 and ± 0.50 -unit of BCS error range between MAN_BCS and AI_BCS was 44.4, 84.6 and 94.8%, respectively, greater than the PA obtained between HA1 vs. HA3. The Bland-Altman plot revealed a minimal systematic bias of -0.09 with a proportional bias at the extreme scores. Furthermore, despite the low κ_w of 0.20, the overall PA at the exact and ± 0.25 -unit of BCS error range between MAN_BCS and AI_BCS regarding the ΔBCS was 45.7 and 88.2%, respectively. A strong linear relationship was observed between BFT and AI_BCS (ρ = 0.75), although weaker than that between BFT and MAN_BCS (ρ = 0.91). The system was able to predict single BCS and ΔBCS with satisfactory accuracy, comparable to that obtained between trained human scorers.

Foot characteristics have been linked to the development of sole lesions (sole hemorrhage and sole ulcers) and white line lesions, also known as claw horn disruption lesions (CHDL). The objective of this study was to examine the association of claw anatomy and sole temperature with the development of CHDL. A cohort of 2,352 cows was prospectively enrolled from 4 UK farms and assessed at 3 time points: before calving (T1precalving), immediately after calving (T2-calving), and in early lactation. At each time point body condition score was recorded, a thermography image of each foot was taken for sole temperature measurement, the presence of CHDL was assessed by veterinary surgeons, and an ultrasound image was taken to retrospectively measure the digital cushion and sole horn thickness. Additionally, at the postcalving time point, foot angle and heel depth were recorded. Four multivariable logistic regression models were fit to separately examine the relationship of precalving and postcalving explanatory variables with the development of either white line lesions or sole lesions. Explanatory variables tested included digital cushion thickness, sole horn thickness, sole temperature, foot angle, and heel depth. Farm, parity, body condition score, and presence of lesion at the time of measurement were also included in the models. A thicker digital cushion shortly after calving was associated with decreased odds of cows developing sole lesions during early lactation (odds ratio [OR]: 0.74, 95% confidence interval [CI]: 0.65-0.84). No association was found between digital cushion thickness and development of white line lesions. Sole temperature after calving was associated with increased odds of the development of sole lesions (OR: 1.03, 95% CI: 1.02-1.05), and sole temperature before and after calving was associated with the development of white line lesions (T1-precalving OR: 1.04, 95% CI: 1.01-1.07; T2-calving OR: 0.96, 95% CI: 0.93-0.99). Neither foot angle nor heel depth was associated with the development of either lesion type. However, an increased sole horn thickness after calving reduced the odds of cows developing sole lesions during early lactation (OR: 0.88, 95% CI: 0.83-0.93), highlighting the importance of maintaining adequate sole horn when foot trimming. Before calving, animals with a lesion at the time of measurement and a thicker sole were more likely to develop a sole lesion (OR: 1.23, 95% CI: 1.09-1.40), compared with those without a sole lesion. The results presented here suggest that white line and sole lesions may have differing etiopathogenesis. Results also confirm the association between the thickness of the digital cushion and the development of sole lesions, highlight the association between sole horn thickness and sole lesions, and challenge the potential importance of foot angle and heel depth in the development of CHDL.

Introduction: Lameness is a major welfare challenge facing the dairy industry worldwide. Monitoring herd lameness prevalence, and early detection and therapeutic intervention are important aspects of lameness control in dairy herds. The objective of this study was to evaluate the performance of a commercially available video surveillance system for automatic detection of dairy cattle lameness (CattleEye Ltd).Methods: This was achieved by first measuring mobility score agreement between CattleEye and two veterinarians (Assessor 1 and Assessor 2), and second, by investigating the ability of the CattleEye system to detect cows with potentially painful foot lesions. We analysed 6,040 mobility scores collected from three dairy farms. Inter-rate agreement was estimated by calculating percentage agreement (PA), Cohen’s kappa (κ) and Gwet’s agreement coefficient (AC). Data regarding the presence of foot lesions were also available for a subset of this dataset. The ability of the system to predict the presence of potentially painful foot lesions was tested against that of Assessor 1 by calculating measures of accuracy, using lesion records during the foot trimming sessions as reference.Results: In general, inter-rater agreement between CattleEye and either human assessor was strong and similar to that between the human assessors, with PA and AC being consistently above 80% and 0.80, respectively. Kappa agreement between CattleEye and the human scorers was in line with previous studies (investigating agreement between human assessors) and within the fair to moderate agreement range. The system was more sensitive than Assessor 1 in identifying cows with potentially painful lesions, with 0.52 sensitivity and 0.81 specificity compared to the Assessor’s 0.29 and 0.89 respectively.Discussion: This pilot study showed that the CattleEye system achieved scores comparable to that of two experienced veterinarians and was more sensitive than a trained veterinarian in detecting painful foot lesions.