Computer vision model for the detection of canine pododermatitis and neoplasia of the paw.

BACKGROUND: Artificial intelligence (AI) has been used successfully in human dermatology. AI utilises convolutional neural networks (CNN) to accomplish tasks such as image classification, object detection and segmentation, facilitating early diagnosis. Computer vision (CV), a field of AI, has shown great results in detecting signs of human skin diseases. Canine paw skin diseases are a common problem in general veterinary practice, and computer vision tools could facilitate the detection and monitoring of disease processes. Currently, no such tool is available in veterinary dermatology. ANIMALS: Digital images of paws from healthy dogs and paws with pododermatitis or neoplasia were used. OBJECTIVES: We tested the novel object detection model Pawgnosis, a Tiny YOLOv4 image analysis model deployed on a microcomputer with a camera for the rapid detection of canine pododermatitis and neoplasia. MATERIALS AND METHODS: The prediction performance metrics used to evaluate the models included mean average precision (mAP), precision, recall, average precision (AP) for accuracy and frames per second (FPS) for speed. RESULTS: A large dataset labelled by a single individual (Dataset A) used to train a Tiny YOLOv4 model provided the best results with a mean mAP of 0.95, precision of 0.86, recall of 0.93 and 20 FPS. CONCLUSIONS AND CLINICAL RELEVANCE: This novel object detection model has the potential for application in the field of veterinary dermatology.

Real-time estimation and forecasting of COVID-19 cases and hospitalizations in Wisconsin HERC regions for public health decision making processes.

BACKGROUND: The spread of the COVID-19 (SARS-CoV-2) and the surging number of cases across the United States have resulted in full hospitals and exhausted health care workers. Limited availability and questionable reliability of the data make outbreak prediction and resource planning difficult. Any estimates or forecasts are subject to high uncertainty and low accuracy to measure such components. The aim of this study is to apply, automate, and assess a Bayesian time series model for the real-time estimation and forecasting of COVID-19 cases and number of hospitalizations in Wisconsin healthcare emergency readiness coalition (HERC) regions. METHODS: This study makes use of the publicly available Wisconsin COVID-19 historical data by county. Cases and effective time-varying reproduction number [Formula: see text] by the HERC region over time are estimated using Bayesian latent variable models. Hospitalizations are estimated by the HERC region over time using a Bayesian regression model. Cases, effective Rt, and hospitalizations are forecasted over a 1-day, 3-day, and 7-day time horizon using the last 28 days of data, and the 20%, 50%, and 90% Bayesian credible intervals of the forecasts are calculated. The frequentist coverage probability is compared to the Bayesian credible level to evaluate performance. RESULTS: For cases and effective [Formula: see text], all three time horizons outperform the three credible levels of the forecast. For hospitalizations, all three time horizons outperform the 20% and 50% credible intervals of the forecast. On the contrary, the 1-day and 3-day periods underperform the 90% credible intervals. Questions about uncertainty quantification should be re-calculated using the frequentist coverage probability of the Bayesian credible interval based on observed data for all three metrics. CONCLUSIONS: We present an approach to automate the real-time estimation and forecasting of cases and hospitalizations and corresponding uncertainty using publicly available data. The models were able to infer short-term trends consistent with reported values at the HERC region level. Additionally, the models were able to accurately forecast and estimate the uncertainty of the measurements. This study can help identify the most affected regions and major outbreaks in the near future. The workflow can be adapted to other geographic regions, states, and even countries where decision-making processes are supported in real-time by the proposed modeling system.

FACTORS IMPACTING SUCCESSFUL REHABILITATION OF BIG BROWN BATS (EPTESICUS FUSCUS) IN A WISCONSIN WILDLIFE REHABILITATION CENTER: A 5-YEAR RETROSPECTIVE.

The big brown bat (Eptesicus fuscus; EPFU) is widely distributed throughout the Americas and plays critical roles in sustaining cave ecosystems and abating agricultural pests. In Wisconsin, EPFU is a threatened species with declining populations due to hibernacula disturbances, wind turbines, and habitat destruction. Due to their ecological and economic value, it is important to be able to release EPFU that enter wildlife rehabilitation centers back to the wild. This study evaluated the medical records of 454 EPFU (275 male, 179 female) admitted to a wildlife rehabilitation center in Wisconsin from 2015 to 2020. For each bat, the season at intake, examination findings, length of time in rehabilitation, and final outcome (released or not released) were recorded. Using a multiple variable logistic regression model, there was a statistically significant positive association between length of time in the rehabilitation center and likelihood of release (odds ratio [OR] 1.08; 95% CI 1.06-1.12); this association can be explained by the need to overwinter some otherwise healthy bats in rehabilitation during hibernation. The following examination findings were associated with a significantly lower likelihood of release: wing injury (OR 0.32; 95% CI 0.10-0.89) and decreased body condition (OR 0.29; 95% CI 0.12-0.64). When corrected for time spent in rehabilitation (potentially artificially lengthened due to hibernation), patients admitted in the summer and fall were less likely to be released than those admitted in the winter (OR 0.93; 95% CI 0.90-0.96 and OR 0.95; 95% CI 0.92-0.97, respectively). The results of this study can be used to help veterinarians and licensed rehabilitators better triage EPFU during admission to wildlife rehabilitation centers in order to improve management and promote successful release back to the wild.

Large-scale genome analysis of bovine commensal Escherichia coli reveals that bovine-adapted E. coli lineages are serving as evolutionary sources of the emergence of human intestinal pathogenic strains.

How pathogens evolve their virulence to humans in nature is a scientific issue of great medical and biological importance. Shiga toxin (Stx)-producing Escherichia coli (STEC) and enteropathogenic E. coli (EPEC) are the major foodborne pathogens that can cause hemolytic uremic syndrome and infantile diarrhea, respectively. The locus of enterocyte effacement (LEE)-encoded type 3 secretion system (T3SS) is the major virulence determinant of EPEC and is also possessed by major STEC lineages. Cattle are thought to be the primary reservoir of STEC and EPEC. However, genome sequences of bovine commensal E. coli are limited, and the emerging process of STEC and EPEC is largely unknown. Here, we performed a large-scale genomic comparison of bovine commensal E. coli with human commensal and clinical strains, including EPEC and STEC, at a global level. The analyses identified two distinct lineages, in which bovine and human commensal strains are enriched, respectively, and revealed that STEC and EPEC strains have emerged in multiple sublineages of the bovine-associated lineage. In addition to the bovine-associated lineage-specific genes, including fimbriae, capsule, and nutrition utilization genes, specific virulence gene communities have been accumulated in stx- and LEE-positive strains, respectively, with notable overlaps of community members. Functional associations of these genes probably confer benefits to these E. coli strains in inhabiting and/or adapting to the bovine intestinal environment and drive their evolution to highly virulent human pathogens under the bovine-adapted genetic background. Our data highlight the importance of large-scale genome sequencing of animal strains in the studies of zoonotic pathogens.

Hot topic: Detecting digital dermatitis with computer vision.

Digital dermatitis (DD) is linked to severe lameness, infertility, and decreased milk production in cattle. Early detection of DD provides an improved prognosis for treatment and recovery; however, this is extremely challenging on commercial dairy farms. Computer vision (COMV) models can help facilitate early DD detection on commercial dairy farms. The aim of this study was to develop and implement a novel COMV tool to identify DD lesions on a commercial dairy farm. Using a database of more than 3,500 DD lesion images, a model was trained using the YOLOv2 architecture to detect the M-stages of DD. The YOLOv2 COMV model detected DD with an accuracy of 71%, and the agreement was quantified as "moderate" by Cohen's kappa when compared with a human evaluator for the internal validation. In the external validation, the YOLOv2 COMV model detected DD with an accuracy of 88% and agreement was quantified as "fair" by Cohen's kappa. Implementation of COMV tools for DD detection provides an opportunity to identify cows for DD treatment, which has the potential to lower DD prevalence and improve animal welfare on commercial dairy farms.

New insights into the association between lameness, behavior, and performance in Simmental cows.

In a herd of 100 milking Simmental cows, data of performance and behavior parameters were collected automatically with different systems such as pedometers, an automatic milking system, and automatic weighing troughs for 1 yr. Performance measures were several milking-related parameters, live weight, as well as feed intake. Behavior-associated measures were feeding behavior (e.g. feeding duration, number of visits to the trough, and feeding pace) as well as activity such as lying duration, number of lying bouts, and overall activity. In the same time, lameness status of every cow was assessed with weekly locomotion scoring. According to the score animals were then classified lame (score 4 or 5) or nonlame (score 1, 2, or 3). From these data in total, 25 parameters summarized to daily values were evaluated for their ability to determine the lameness status of a cow. Data were analyzed with a regularized regression method called elastic net with the outcome lame or nonlame. The final model had a high prediction accuracy with an area under the curve of 0.91 [95% confidence interval (CI) = 0.88-0.94]. Specificity was 0.81 (95% CI = 0.73-0.85) and sensitivity was 0.94 (95% CI = 0.88-1.00). The most important factors associated with a cow being lame were number of meals, average feed intake per meal, and average duration of a meal. Lame cows fed in fewer and shorter meals with a decreased intake per meal. Milk yield and lying-behavior-associated parameters were relevant in the model, too, but only as parts of interaction terms demonstrating their strong dependence on other factors. A higher milk yield only resulted in higher risk of being lame if feed intake was decreased. The same accounts for lying duration: only if lying time was below the 50% quantile did an increased milk yield result in a higher risk of being lame. The association of lameness and daily lying duration was influenced by daily feeding duration and feeding duration at daytime. The results of the study give deeper insights on how the association between behavior and performance parameters and lameness is influenced by intrinsic factors in particular and that many of these have to be considered when trying to predict lameness based on such data. The findings lead to a better understanding why, for instance, lying duration or milk yield seem to be highly correlated with lameness in cows but still have not been overly useful as parameters in other lameness detection models.

Acquisition and persistence of strain-specific methicillin-resistant Staphylococcus aureus and their determinants in community nursing homes.

BACKGROUND: Nursing home residents are frequently colonized with various strains of methicillin-resistant Staphylococcus aureus (MRSA) but the intra-facility dynamics of strain-specific MRSA remains poorly understood. We aimed at identifying and quantifying the associations between acquisition and carriage of MRSA strains and their potential risk factors in community nursing homes using mathematical modeling. METHODS: The data was collected during a longitudinal MRSA surveillance study in six nursing homes in South Central Wisconsin. MRSA cultures were obtained from subjects every 3 months for up to one year. MRSA isolates were subsequently strain-typed by pulsed-field gel electrophoresis (PFGE), and their genetic similarity was established based on the Dice coefficients. Bayesian network analysis, logistic regression and elastic net were used to quantify the associations between acquisition and carriage of MRSA strains discriminated at 80% and 95% strain similarity thresholds and potentially modifiable resident characteristics including previous antibiotic exposure, comorbidity, medical devices, chronic wounds, functional and cognitive status and recent hospitalizations. RESULTS: Absence of severe cognitive impairment as well as presence of a wound, device and severe comorbidity was associated with elevated probability of USA100 carriage although there was a variation based on the combination of those risk factors. Residents with severe comorbidity and cognitive status and presence of device and wound were identified as certain carriers of USA100 in our sample. Residents with a chronic wound were more likely to carry USA100 MRSA (OR = 2.77, 95% CI = 1.37-5.87). Functional status was identified as an important determinant of carriage of USA100 and USA300 strains. Comorbidity and cognitive status were the two factors associated with carriage of all clonal groups in the study (USA100, USA300 and USA1200). CONCLUSIONS: The combination of Bayesian network analysis, logistic regression and elastic net can be used to identify associations between acquisition and carriage of MRSA strains and their potential risk factors in the face of scarce data. The revealed associations may be used to generate hypothesis for further study of determinants of acquisition and carriage of selected MRSA subtypes and to better inform infection control efforts in community nursing homes.

Evaluating the efficacy of beef slaughter line interventions by quantifying the six major non-O157 Shiga toxin producing Escherichia coli serogroups using real-time multiplex PCR.

Six major Shiga toxin producing Escherichia coli (STEC) serogroups: O26, O103, O145, O111, O121, and O45 have been declared as adulterants in federally inspected raw beef in the USA effective June 4th, 2012 in addition to the routinely tested STEC O157: H7. This study tests a real-time multiplex PCR assay and pooling of the samples to optimize the detection and quantification (prevalence and contamination) of six major non-O157 STEC, regardless of possessing Shiga toxins. To demonstrate the practicality, one large-scale slaughter plant (Plant LS) and one small-scale slaughter plant (Plant SS) located in the Mid-Western USA were sampled, in 2011, before the establishment of 2013 USDA laboratory protocols. Carcasses were sampled at consecutive intervention stations and beef trimmings were collected at the end of the fabrication process. Plant SS had marginally more contaminated samples than Plant LS (p-value 0.08). The post-hide removal wash, steam pasteurization, and lactic acid (≤5%) spray used in Plant LS seemed to reduce the six serogroups effectively, compared to the hot-water wash and 7-day chilling at Plant SS. Compared to the culture isolation methods, quantification of the non-O157 STEC using real-time PCR may be an efficient way to monitor the efficacy of slaughter line interventions.

Evaluation of the Use of Zero-Augmented Regression Techniques to Model Incidence of Campylobacter Infections in FoodNet.

The Foodborne Diseases Active Surveillance Network (FoodNet) is currently using a negative binomial (NB) regression model to estimate temporal changes in the incidence of Campylobacter infection. FoodNet active surveillance in 483 counties collected data on 40,212 Campylobacter cases between years 2004 and 2011. We explored models that disaggregated these data to allow us to account for demographic, geographic, and seasonal factors when examining changes in incidence of Campylobacter infection. We hypothesized that modeling structural zeros and including demographic variables would increase the fit of FoodNet's Campylobacter incidence regression models. Five different models were compared: NB without demographic covariates, NB with demographic covariates, hurdle NB with covariates in the count component only, hurdle NB with covariates in both zero and count components, and zero-inflated NB with covariates in the count component only. Of the models evaluated, the nonzero-augmented NB model with demographic variables provided the best fit. Results suggest that even though zero inflation was not present at this level, individualizing the level of aggregation and using different model structures and predictors per site might be required to correctly distinguish between structural and observational zeros and account for risk factors that vary geographically.

Customized recommendations for production management clusters of North American automatic milking systems.

Automatic milking systems (AMS) are implemented in a variety of situations and environments. Consequently, there is a need to characterize individual farming practices and regional challenges to streamline management advice and objectives for producers. Benchmarking is often used in the dairy industry to compare farms by computing percentile ranks of the production values of groups of farms. Grouping for conventional benchmarking is commonly limited to the use of a few factors such as farms' geographic region or breed of cattle. We hypothesized that herds' production data and management information could be clustered in a meaningful way using cluster analysis and that this clustering approach would yield better peer groups of farms than benchmarking methods based on criteria such as country, region, breed, or breed and region. By applying mixed latent-class model-based cluster analysis to 529 North American AMS dairy farms with respect to 18 significant risk factors, 6 clusters were identified. Each cluster (i.e., peer group) represented unique management styles, challenges, and production patterns. When compared with peer groups based on criteria similar to the conventional benchmarking standards, the 6 clusters better predicted milk produced (kilograms) per robot per day. Each cluster represented a unique management and production pattern that requires specialized advice. For example, cluster 1 farms were those that recently installed AMS robots, whereas cluster 3 farms (the most northern farms) fed high amounts of concentrates through the robot to compensate for low-energy feed in the bunk. In addition to general recommendations for farms within a cluster, individual farms can generate their own specific goals by comparing themselves to farms within their cluster. This is very comparable to benchmarking but adds the specific characteristics of the peer group, resulting in better farm management advice. The improvement that cluster analysis allows for is characterized by the multivariable approach and the fact that comparisons between production units can be accomplished within a cluster and between clusters as a choice.

Factors associated with increased milk production for automatic milking systems.

Automatic milking systems (AMS) are increasingly popular throughout the world. Our objective was to analyze 635 North American dairy farms with AMS for (risk) factors associated with increased milk production per cow per day and milk production per robot per day. We used multivariable generalized mixed linear regressions, which identified several significant risk factors and interactions of risk factors associated with milk production. Free traffic was associated with increased production per cow and per robot per day compared with forced systems, and the presence of a single robot per pen was associated with decreased production per robot per day compared with pens using 2 robots. Retrofitted farms had significantly less production in the first 4 yr since installation compared with production after 4 yr of installation. In contrast, newly built farms did not see a significant change in production over time since installation. Overall, retrofitted farms did not produce significantly more or less milk than newly constructed farms. Detailed knowledge of factors associated with increased production of AMS will help guide future recommendations to producers looking to transition to an AMS and maximize their production.

World Health Organization Estimates of the Global and Regional Disease Burden of 22 Foodborne Bacterial, Protozoal, and Viral Diseases, 2010: A Data Synthesis.

BACKGROUND: Foodborne diseases are important worldwide, resulting in considerable morbidity and mortality. To our knowledge, we present the first global and regional estimates of the disease burden of the most important foodborne bacterial, protozoal, and viral diseases. METHODS AND FINDINGS: We synthesized data on the number of foodborne illnesses, sequelae, deaths, and Disability Adjusted Life Years (DALYs), for all diseases with sufficient data to support global and regional estimates, by age and region. The data sources included varied by pathogen and included systematic reviews, cohort studies, surveillance studies and other burden of disease assessments. We sought relevant data circa 2010, and included sources from 1990-2012. The number of studies per pathogen ranged from as few as 5 studies for bacterial intoxications through to 494 studies for diarrheal pathogens. To estimate mortality for Mycobacterium bovis infections and morbidity and mortality for invasive non-typhoidal Salmonella enterica infections, we excluded cases attributed to HIV infection. We excluded stillbirths in our estimates. We estimate that the 22 diseases included in our study resulted in two billion (95% uncertainty interval [UI] 1.5-2.9 billion) cases, over one million (95% UI 0.89-1.4 million) deaths, and 78.7 million (95% UI 65.0-97.7 million) DALYs in 2010. To estimate the burden due to contaminated food, we then applied proportions of infections that were estimated to be foodborne from a global expert elicitation. Waterborne transmission of disease was not included. We estimate that 29% (95% UI 23-36%) of cases caused by diseases in our study, or 582 million (95% UI 401-922 million), were transmitted by contaminated food, resulting in 25.2 million (95% UI 17.5-37.0 million) DALYs. Norovirus was the leading cause of foodborne illness causing 125 million (95% UI 70-251 million) cases, while Campylobacter spp. caused 96 million (95% UI 52-177 million) foodborne illnesses. Of all foodborne diseases, diarrheal and invasive infections due to non-typhoidal S. enterica infections resulted in the highest burden, causing 4.07 million (95% UI 2.49-6.27 million) DALYs. Regionally, DALYs per 100,000 population were highest in the African region followed by the South East Asian region. Considerable burden of foodborne disease is borne by children less than five years of age. Major limitations of our study include data gaps, particularly in middle- and high-mortality countries, and uncertainty around the proportion of diseases that were foodborne. CONCLUSIONS: Foodborne diseases result in a large disease burden, particularly in children. Although it is known that diarrheal diseases are a major burden in children, we have demonstrated for the first time the importance of contaminated food as a cause. There is a need to focus food safety interventions on preventing foodborne diseases, particularly in low- and middle-income settings.

Data-driven methods for imputing national-level incidence in global burden of disease studies.

OBJECTIVE: To develop transparent and reproducible methods for imputing missing data on disease incidence at national-level for the year 2005. METHODS: We compared several models for imputing missing country-level incidence rates for two foodborne diseases - congenital toxoplasmosis and aflatoxin-related hepatocellular carcinoma. Missing values were assumed to be missing at random. Predictor variables were selected using least absolute shrinkage and selection operator regression. We compared the predictive performance of naive extrapolation approaches and Bayesian random and mixed-effects regression models. Leave-one-out cross-validation was used to evaluate model accuracy. FINDINGS: The predictive accuracy of the Bayesian mixed-effects models was significantly better than that of the naive extrapolation method for one of the two disease models. However, Bayesian mixed-effects models produced wider prediction intervals for both data sets. CONCLUSION: Several approaches are available for imputing missing data at national level. Strengths of a hierarchical regression approach for this type of task are the ability to derive estimates from other similar countries, transparency, computational efficiency and ease of interpretation. The inclusion of informative covariates may improve model performance, but results should be appraised carefully.

Short communication: Determination of the ability of Thymox to kill or inhibit various species of microorganisms associated with infectious causes of bovine lameness in vitro.

Infectious claw diseases continue to plague cattle in intensively managed husbandry systems. Poor foot hygiene and constant moist environments lead to the infection and spread of diseases such as digital dermatitis (hairy heel warts), interdigital dermatitis, and interdigital phlegmon (foot rot). Currently, copper sulfate and formalin are the most widely used disinfecting agents in bovine footbaths; however, the industry could benefit from more environmentally and worker friendly substitutes. This study determined the in vitro minimum inhibitory concentrations and minimum bactericidal concentrations of Thymox (Laboratoire M2, Sherbrooke, Québec, Canada) for a selection of microorganisms related to infectious bovine foot diseases. Thymox is a broad-spectrum agricultural disinfectant that is nontoxic, noncorrosive, and readily biodegradable. The values for minimum inhibitory concentration and minimum bactericidal concentration indicated that Thymox inhibited growth and killed the various species of microorganisms under study at much lower concentrations compared with the recommended working concentration of a 1% solution. Overall, the values found in this study of minimum inhibitory concentration and minimum bactericidal concentration of Thymox show its potential as an alternative antibacterial agent used in bovine footbaths; however, field trials are needed to determine its effectiveness for the control and prevention of infectious claw diseases.

Musculoskeletal lesions and lameness in 121 horses with carpal sheath effusion (1999-2010).

Equine carpal sheath effusion has multiple etiologies. The purpose of this retrospective study was to describe the prevalence of distinct musculoskeletal lesions lameness in a sample of horses with a clinical diagnosis of carpal sheath effusion. A total of 121 horses met inclusion criteria. Seventy-four percent (89/121) of horses were lame at presentation; middle-aged (9-18 years, 80%) and older (> 18 years, 85%) horses were lame more frequently than young horses (< 9 years, 44%). Ninety-three percent (113/121) were diagnosed with osseous and/or soft tissue abnormalities. Of these 113 horses, 10 exhibited osseous abnormalities, whereas 111 were diagnosed with soft tissue lesions. Eighty-four percent (93/111) of the soft tissue injuries extended from the caudodistal antebrachium to the palmar metacarpus. The superficial digital flexor tendon (98/111; 88%) and accessory ligament of the superficial digital flexor tendon (64/111; 58%) were the most commonly injured structures, with both structures affected in 41 (41/111; 37%) horses. Injuries within the caudodistal antebrachium included the superficial digital flexor musculotendinous junction (66), the accessory ligament of the superficial digital flexor tendon (64), and deep digital flexor muscle (21), in isolation or in combination with other structures. Increased echogenicity in the medial superficial digital flexor musculotendinous junction was detected in 40 horses and was significantly associated with increasing age (middle-aged, 19/40; old, 18/40). Findings from this study indicated that age should be taken into consideration for horses presented with carpal sheath effusion and that adjacent structures within the caudodistal antebrachium should be included in evaluations.

Phylogenetic characterization of Escherichia coli O157 : H7 based on IS629 distribution and Shiga toxin genotype.

Shiga toxin (stx)-producing Escherichia coli O157 : H7 is a prominent food-borne pathogen. Symptoms in human infections range from asymptomatic to haemorrhagic colitis and haemolytic uraemic syndrome, and there is a need for methods that yield information that can be used to better predict clinical and epidemiological outcomes. IS629 is an insertion sequence notable for its prevalence and variable distribution in the chromosome of E. coli O157 : H7, which has been exploited for subtyping and strain characterization. In particular, IS629 distribution is closely aligned with the major phylogenetic lineages that are known to be distinctive in their genome structure and virulence potential. In the present study, a comprehensive subtyping method in which IS629-typing was combined with stx genotyping was developed using a conventional PCR approach. This method consisted of a set of 32 markers based on the unique distribution of IS629 in the three major phylogenetic lineages of E. coli O157 : H7 and six additional markers to determine the stx genotype, a key virulence signature associated with each lineage. The analysis of IS629 loci variation with the 32 markers allowed us to determine the IS629 distribution profile (IDP), phylogenetic lineage and genetic relatedness of the 31 E. coli O157 : H7 strains examined. An association between IDP typing and stx genotype was observed. The use of both IDP and the stx genotype for strain characterization provided confirmative and complementary data in support of lineage placement of closely related strains. In addition, IS629/stx profiles were in agreement with strain segregation based on LSPA-6 (lineage-specific polymorphism assay) and PFGE subtyping, demonstrating its potential as a subtyping and strain tracking method.

Benchmarking analysis of computer vision algorithms on edge devices for the real-time detection of digital dermatitis in dairy cows.

Digital dermatitis (DD) is a bovine claw disease responsible for ulcerative lesions on the coronary band of the foot. It causes significant animal welfare and economic losses to the cattle industry. Early detection of DD can lead to prompt treatment and decrease lameness. Current detection and staging methods require a trained individual to evaluate the interdigital space on each foot for clinical signs of DD. Computer vision (CV), a type of artificial intelligence for image analysis, has demonstrated promising results on object detection tasks. However, farms require robust solutions that can be deployed in harsh conditions including dust, debris, humidity, precipitation, other equipment issues. The study aims to train, deploy, and benchmark DD detection models on edge devices. Images were collected from commercial dairy farms with the camera facing the interdigital space on the plantar surface of the foot. Images were scored for M-stages of DD by a trained investigator using the M-stage DD classification system. Models were trained to detect and score DD lesions and embedded on an edge device. The Tiny YOLOv4 model deployed on a CV specific integrated camera module connected to a single board computer achieved a mean average precision (mAP) of 0.895, an overall prediction accuracy of 0.873, and a Cohen's kappa of 0.830 for agreement between the computer vision model and the trained investigator. The model reached a final inference speed of 40 frames per second (FPS) and ran stably without any interruptions. The CV model was able to detect DD lesions on an edge device with high performance and speed. The CV tool can be used for early detection and prompt treatment of DD in dairy cows. Real-time detection of DD on edge device will improve health outcomes, while simultaneously decreasing labor costs. We demonstrate that the deployed model can be a low-power and portable solution for real-time detection of DD on dairy farms. This result is a step towards applying CV algorithms to veterinary medicine and implementing real-time detection of health outcomes in precision farming.

DairyCoPilot-Automated data compilation and analysis tools for DairyComp data assets.

Modern dairy farm management requires meaningful data and careful analysis to maximize profitability, cow health, and welfare. Current data platforms, such as DairyComp, lack robust integrated data analysis tools. Producers and consultants need dedicated tools to turn collected data sets into assets for informed decision-making processes. The DairyCoPilot app allows users to rapidly extract health and production data from DairyComp, then compile and analyze the data using a menu-driven point-and-click approach. Prospects for training consultants in applied data analysis skills make DairyCoPilot a tool to identify farm management bottlenecks with less time spent for data analysis, improving cow health, and dairy production. The DairyCoPilot Dashboard R Shiny application is published using RStudio Connect: https://connect.doit.wisc.edu/dairy-copilot/.

Comparative analysis of computer vision algorithms for the real-time detection of digital dermatitis in dairy cows.

Digital dermatitis (DD) is a bovine claw disease responsible for ulcerative lesions on the planar aspect of the hoof. DD is associated with massive herd outbreaks of lameness and influences cattle welfare and production. Early detection of DD can lead to prompt treatment and decrease lameness. Computer vision (CV) provides a unique opportunity to improve early detection. The study aims to train and compare applications for the real-time detection of DD in dairy cows. Eight CV models were trained for detection and scoring, compared using performance metrics and inference time, and the best model was automated for real-time detection using images and video. Images were collected from commercial dairy farms while facing the interdigital space on the plantar surface of the foot. Images were scored for M-stages of DD by a trained investigator using the M-stage DD classification system with distinct labels for hyperkeratosis (H) and proliferations (P). Two sets of images were compiled: the first dataset (Dataset 1) containing 1,177 M0/M4H and 1,050 M2/M2P images and the second dataset (Dataset 2) containing 240 M0, 17 M2, 51 M2P, 114 M4H, and 108 M4P images. Models were trained to detect and score DD lesions and compared for precision, recall, and mean average precision (mAP) in addition to inference time in frame per second (FPS). Seven of the nine CV models performed well compared to the ground truth of labeled images using Dataset 1. The six models, Faster R-CNN, Cascade R-CNN, YOLOv3, Tiny YOLOv3, YOLOv4, Tiny YOLOv4, and YOLOv5s achieved an mAP between 0.964 and 0.998, whereas the other two models, SSD and SSD Lite, yielded an mAP of 0.371 and 0.387 respectively. Overall, YOLOv4, Tiny YOLOv4, and YOLOv5s outperformed all other models with almost perfect precision, perfect recall, and a higher mAP. Tiny YOLOv4 outperformed all other models with respect to inference time at 333 FPS, followed by YOLOv5s at 133 FPS and YOLOv4 at 65 FPS. YOLOv4 and Tiny YOLOv4 performed better than YOLOv5s compared to the ground truth using Dataset 2. YOLOv4 and Tiny YOLOv4 yielded a similar mAP of 0.896 and 0.895, respectively. However, Tiny YOLOv4 achieved both higher precision and recall compared to YOLOv4. Finally, Tiny YOLOv4 was able to detect DD lesions on a commercial dairy farm with high performance and speed. The proposed CV tool can be used for early detection and prompt treatment of DD in dairy cows. This result is a step towards applying CV algorithms to veterinary medicine and implementing real-time DD detection on dairy farms.

Evolution of the Stx2-encoding prophage in persistent bovine Escherichia coli O157:H7 strains.

Escherichia coli O157:H7 is a human pathogen that resides asymptomatically in its bovine host. The level of Shiga toxin (Stx) produced is variable in bovine-derived strains in contrast to human isolates that mostly produce high levels of Stx. To understand the genetic basis for varied Stx production, chronological collections of bovine isolates from Wisconsin dairy farms, R and X, were analyzed for multilocus prophage polymorphisms, stx(2) subtypes, and the levels of stx(2) transcript and toxin. The E. coli O157:H7 that persisted on both farms were phylogenetically distinct and yet produced little to no Stx2 due to gene deletions in Stx2c-encoding prophage (farm R) or insertional inactivation of stx(2a) by IS1203v (farm X). Loss of key regulatory and lysis genes in Stx2c-encoding prophage abolished stx(2c) transcription and induction of the prophage and stx(2a)::IS1203v in Stx2a-encoding prophage generated a truncated stx(2a) mRNA without affecting phage production. Stx2-producing strains were transiently present (farm R) and became Stx2 negative on farm X (i.e., stx(2a)::IS1203v). To our knowledge, this is the first study that details the evolution of E. coli O157:H7 and its Stx2-encoding prophage in a chronological collection of natural isolates. The data suggest the bovine and farm environments can be niches where Stx2-negative E. coli O157:H7 emerge and persist, which explains the Stx variability in bovine isolates and may be part of an evolutionary step toward becoming bovine specialists.