Evaluation of a gait scoring system for cattle by using cluster analysis and Krippendorff's α reliability

Automated dairy cattle lameness detection utilizing the power of artificial intelligence; current status quo and future research opportunities

Lameness represents a major welfare and health problem for the dairy industry across all farming systems. Visual mobility scoring, although very useful, is labour-intensive and physically demanding, especially in large dairies, often leading to inconsistencies and inadequate uptake of the practice. Technological and computational advancements of artificial intelligence (AI) have led to the development of numerous automated solutions for livestock monitoring. The objective of this study was to review the automated systems using AI algorithms for lameness detection developed to-date. These systems rely on gait analysis using accelerometers, weighing platforms, acoustic analysis, radar sensors and computer vision technology. The lameness features of interest, the AI techniques used to process the data as well as the ground truth of lameness selected in each case are described. Measures of accuracy regarding correct classification of cows as lame or non-lame varied with most systems being able to classify cows with adequate reliability. Most studies used visual mobility scoring as the ground truth for comparison with only a few studies using the presence of specific foot pathologies. Given the capabilities of AI, and the benefits of early treatment of lameness, longitudinal studies to identify gait abnormalities using automated scores related to the early developmental stages of different foot pathologies are required. Farm-specific optimal thresholds for early intervention should then be identified to ameliorate cow health and welfare but also minimise unnecessary inspections.

Accuracy and inter-rater agreement among practitioners using quick thoracic ultrasonography to diagnose calf pneumonia

BACKGROUND

Thoracic ultrasonography (TUS) is a commonly used tool for on-farm detection of pneumonia in calves. Different scanning methods have been described, but the performance of novice practitioners after training has not been documented.

METHODS

In this study, 38 practitioners performed quick TUS (qTUS) on 18-23 calves each. Pneumonia was defined as lung consolidation 1 cm or more in depth. Diagnostic parameters (accuracy [Acc], sensitivity [Se] and specificity [Sp]) were compared to those of an experienced operator. Cohen's kappa and Krippendorff's alpha (Kalpha) were determined. The potential effects of training and exam sessions on performance were evaluated.

RESULTS

The average relative Se and Sp were 0.66 (standard deviation [SD] = 0.26; minimum [Min.]-Maximum [Max.] = 0-1) and 0.71 (SD = 0.19; Min.-Max. = 0.25-1), respectively. The average relative Acc was 0.73 (SD = 0.11; Min.-Max. = 0.52-0.96). Over all sessions, Cohen's kappa averaged 0.40 (SD = 0.24; Min.-Max. = 0.014-0.90) and Kalpha was 0.24 (95% confidence interval [CI]: 0.20-0.27), indicating 'fair' agreement. Calf age and housing influenced Se and Sp. Supervised practical training improved Se by 17.5% (95% CI: 0.01-0.34).

LIMITATIONS

The separate effects of calf age and housing could not be determined.

CONCLUSION

This study showed that qTUS, like any other clinical skill, has a learning curve, and variability in performance can be substantial. Adequate training and certification of one's skill are recommended to assure good diagnostic accuracy.

Identification of temporal patterns of environmental heat stress of Holstein dairy heifers raised in Mediterranean climate during their in-utero and post-natal life periods and modelling their effects on age at first calving

A retrospective study was conducted to evaluate temporal patterns of environmental heat stress during the in-uterus period of development (IUP) and the 3-month post-natal (PN) period of dairy heifers, and to estimate their association with the age at first calving (AFC). Data from 30 dairy herds in Northern Greece including 9098 heifers were extracted from National Cattle Database. Data (2005-2019) regarding 230,100 farm-specific ambient daily temperature and relative humidity records, were obtained from ERA5-Land. Average monthly Temperature-Humidity-Index values (THI; low≤68, and high>68) were calculated and matched for each heifer to their IUP and PN. Subsequently, Cluster Analysis was used with monthly THIs as predictors to allocate heifers to THI clusters. The association of clusters with AFC was assessed with Generalized Linear Mixed Model analysis, an extended form of multiple linear regression. Finally, 8 Heat Stress Clusters (HSC; namely HSC-1 to HSC-8) were identified. Compared to HSC-8 (8th-9th IUP months and 1st PN month) heifers of HSC-5 (4th-7th IUP months) and HSC-6 (6th-8th IUP months) calved 13.8 and 17.8 days later, respectively (P < 0.01-0.001). Moreover, when AFC was treated as a binary variable, heifers of HSC-5 and HSC-6 had 1.15 and 1.34 (P < 0.01-0.001) higher risk of calving for the first time later than 787 days compared to HSC-8, respectively.

Assessing the Quality of Mobile Health-Related Apps: Interrater Reliability Study of Two Guides

Background

There is a huge number of health-related apps available, and the numbers are growing fast. However, many of them have been developed without any kind of quality control. In an attempt to contribute to the development of high-quality apps and enable existing apps to be assessed, several guides have been developed.

Objective

The main aim of this study was to study the interrater reliability of a new guide — the Mobile App Development and Assessment Guide (MAG) — and compare it with one of the most used guides in the field, the Mobile App Rating Scale (MARS). Moreover, we also focused on whether the interrater reliability of the measures is consistent across multiple types of apps and stakeholders.

Methods

In order to study the interrater reliability of the MAG and MARS, we evaluated the 4 most downloaded health apps for chronic health conditions in the medical category of IOS and Android devices (ie, App Store and Google Play). A group of 8 reviewers, representative of individuals that would be most knowledgeable and interested in the use and development of health-related apps and including different types of stakeholders such as clinical researchers, engineers, health care professionals, and end users as potential patients, independently evaluated the quality of the apps using the MAG and MARS. We calculated the Krippendorff alpha for every category in the 2 guides, for each type of reviewer and every app, separately and combined, to study the interrater reliability.

Results

Only a few categories of the MAG and MARS demonstrated a high interrater reliability. Although the MAG was found to be superior, there was considerable variation in the scores between the different types of reviewers. The categories with the highest interrater reliability in MAG were “Security” (α=0.78) and “Privacy” (α=0.73). In addition, 2 other categories, “Usability” and “Safety,” were very close to compliance (health care professionals: α=0.62 and 0.61, respectively). The total interrater reliability of the MAG (ie, for all categories) was 0.45, whereas the total interrater reliability of the MARS was 0.29.

Conclusions

This study shows that some categories of MAG have significant interrater reliability. Importantly, the data show that the MAG scores are better than the ones provided by the MARS, which is the most commonly used guide in the area. However, there is great variability in the responses, which seems to be associated with subjective interpretation by the reviewers.

On-farm detection of claw lesions in dairy cows based on acoustic analyses and machine learning

Claw lesions are a serious problem on dairy farms, affecting both the health and welfare of the cow. Automated detection of lameness with a practical, on-farm application would support the early detection and treatment of lame cows, potentially reducing the number and severity of claw lesions. Therefore, in this study, a method was proposed for the detection of claw lesions based on the acoustic analysis of a cow's gait. A panel was constructed to measure the impact sound of animals walking over it. The recorded impact sound was edited, and 640 sound files from 64 cows were analyzed. The classification of animal-lameness status was performed using a machine-learning process with a random forest algorithm. The gold standard was a 2-point scale of hoof-trimming results (healthy vs. affected), and 38 properties of the recorded sound files were used as influencing factors. A prediction model for classifying the cow lameness was built using a random forest algorithm. This was validated by comparing the reference output from hoof-trimming with the model output concerning the impact sound. Altering the likelihood settings and changing the cutoff value to predict lame animals improved the prediction model. At a cutoff at 0.4, a decreased false-negative rate was generated, and the false-positive rate only increased slightly. This model obtained a sensitivity of 0.81 and a specificity of 0.97. With this procedure, Cohen's Kappa value of 0.80 showed good agreement between model classification and diagnoses from hoof-trimming. In summary, the prediction model enabled the detection of cows with claw lesions. This study shows that lameness can be detected by machine learning from the impact sound of hoofs in dairy cows.

Subclinical hypocalcaemia follows specific time-related and severity patterns in post-partum Holstein cows

Hypocalcaemia is a common metabolic disorder of post-parturient dairy cows; enhancement of our knowledge regarding Ca dynamics would improve cow health and welfare. This study investigated the presence of time- and severity-related subclinical hypocalcaemia (SCHCa) patterns in Holstein cows after calving and their association with diseases during the first week of lactation. A total of 938 cows from nine herds were blood sampled on days 1, 2, 4 and 8 post-partum, and serum Ca concentration was measured with atomic absorption. Subclinical ketosis (serum β-hydroxybutyrate≥1.2 mmol/L) and cases of retained foetal membranes, metritis (MET), mastitis, ketosis and displaced abomasum (DA) were recorded. Using receiver operating characteristic (ROC) analysis, the SCHCa cut-off was identified at serum Ca ≤ 2.09 mmol/L. Hierarchical and two-step cluster analysis classified cows to seven clusters based on test day cow Ca records and overall SCHCa status. Two clusters (NORM and HIGH) included 318 normocalcaemic cows and five clusters (SCH-1 to -5) those that were at least once subclinically hypocalcaemic (n = 620). A second ROC analysis was performed in order to distinguish mild from severe cases of SCHCa in these 620 cows; this cut-off was identified at 1.93 mmol/L. The associated risk of disease with Ca clustership was assessed with generalized linear mixed models. Overall incidence of SCHCa was 66.1%. Clusters SCH-1 and SCH-2 included cows with short-term SCHCa of day 1 and day 2, respectively, while SCH-3 included cows with severe and prolonged SCHCa extending to day 4 and beyond. Clusters SCH-4 and SCH-5 included cows with delayed SCHCa, which culminated on days 4 and 8, respectively. Compared to NORM cows in HIGH had lower risk of MET and no cases of DA. Cows in SCH-3 had higher risk of being diagnosed with retained foetal membranes, DA or any disease during the study period. Cows in SCH-5 had higher risk of being diagnosed with ketosis, subclinical ketosis or any disease. In conclusion, there are multiple normocalcaemic and hypocalcaemic patterns that are differently associated with disease risk.

Using the Footfall Sound of Dairy Cows for Detecting Claw Lesions

Simple Summary

Claw lesions affect the health and well-being of dairy cows and are the third most frequent reason for culling. The most common method to identify cows affected by claw lesions is the assessment of locomotion. However, this requires apart from a trained observer a lot of time. And time is an important limiting factor especially with increasing herd sizes. Thus, regarding the detection of claw lesions, a sensor-assisted, automated system to be implemented on farms should be developed. This system analyzes the footfall sound of cows to distinguish lame from non-lame animals. Using a runway with sensors for recording the footfall sound, we found that cows with non-infectious diseases such as sole ulcers showed a less forceful gait pattern than healthy ones. It is known that particular cows with non-infectious diseases have a greater sensitivity to pain. Therefore the established system is capable of selecting cows that take more cautious steps. This character of abnormal gait pattern is in turn a sign of lameness. Therefore, the automated system as developed in this study is a promising tool for detecting lameness in dairy cows on farms.

Abstract

An important factor for animal welfare in cattle farming is the detection of lameness. The presented study is part of a project aiming to develop a system that is capable of an automated diagnosis of claw lesions by analyzing the footfall sound. Data were generated from cows walking along a measurement zone where piezoelectric sensors recorded their footfall sounds. Locomotion of the animals was scored and they were graded according to a three-scale scoring system (LS1 = non-lame; LS2 = uneven gait; LS3 = lame). Subsequently, the cows were examined by a hoof trimmer. The walking speed across the test track was significantly higher in cows with LS1 compared to those with LS2 and LS3 and thus, they were showing a smoother gait pattern. The standard deviation of volume (SDV) in the recorded footfall sound signal was considered as a factor for the force of a cow’s footsteps. Cows with non-infectious claw lesions showed lower SDV than healthy cows and those with infectious claw diseases. This outcome confirmed the hypothesis that the evaluated cows affected by non-infectious claw lesions have a greater sensitivity to pain and demonstrate a less forceful gait pattern. These first results clearly show the potential of using footfall sound analysis for detecting claw lesions.