A New Method to Detect Buffalo Mastitis Using Udder Ultrasonography Based on Deep Learning Network

Mastitis is one of the most predominant diseases with a negative impact on ranch products worldwide. It reduces milk production, damages milk quality, increases treatment costs, and even leads to the premature elimination of animals. In addition, failure to take effective measures in time will lead to widespread disease. The key to reducing the losses caused by mastitis lies in the early detection of the disease. The application of deep learning with powerful feature extraction capability in the medical field is receiving increasing attention. The main purpose of this study was to establish a deep learning network for buffalo quarter-level mastitis detection based on 3054 ultrasound images of udders from 271 buffaloes. Two data sets were generated with thresholds of somatic cell count (SCC) set as 2 × 105 cells/mL and 4 × 105 cells/mL, respectively. The udders with SCCs less than the threshold value were defined as healthy udders, and otherwise as mastitis-stricken udders. A total of 3054 udder ultrasound images were randomly divided into a training set (70%), a validation set (15%), and a test set (15%). We used the EfficientNet_b3 model with powerful learning capabilities in combination with the convolutional block attention module (CBAM) to train the mastitis detection model. To solve the problem of sample category imbalance, the PolyLoss module was used as the loss function. The training set and validation set were used to develop the mastitis detection model, and the test set was used to evaluate the network's performance. The results showed that, when the SCC threshold was 2 × 105 cells/mL, our established network exhibited an accuracy of 70.02%, a specificity of 77.93%, a sensitivity of 63.11%, and an area under the receiver operating characteristics curve (AUC) of 0.77 on the test set. The classification effect of the model was better when the SCC threshold was 4 × 105 cells/mL than when the SCC threshold was 2 × 105 cells/mL. Therefore, when SCC ≥ 4 × 105 cells/mL was defined as mastitis, our established deep neural network was determined as the most suitable model for farm on-site mastitis detection, and this network model exhibited an accuracy of 75.93%, a specificity of 80.23%, a sensitivity of 70.35%, and AUC 0.83 on the test set. This study established a 1/4 level mastitis detection model which provides a theoretical basis for mastitis detection in buffaloes mostly raised by small farmers lacking mastitis diagnostic conditions in developing countries.

Pattern of milk yield and immunoglobulin concentration and factors associated with colostrum quality at the quarter level in dairy cows after parturition

First colostrum yield and constituents as well as milk yield during established lactation vary considerably among mammary quarters in dairy cows. However, data on the development of milk yield, IgG concentration, and their distribution per quarter within cows during the first milkings after calving are scarce. We analyzed milk production and IgG concentration at the individual quarter level in 29 multiparous Holstein cows during the first 5 milkings after calving. Cow- and calf-related factors (time interval between calving and first milking, parity number, previous lactation yield, gestation length, dry period length, sex, and birth weight of the calf) potentially affecting first colostrum quality and quantity were assessed. Milking of first colostrum was carried out between 30 and 180 min after parturition. Further milkings were performed twice daily. Quarter milk yield varied between 0.1 and 5.5 kg at the first milking and between 1.4 and 5.1 kg at the fifth milking relative to parturition. Quarter IgG concentration ranged between 18.8 and 106.0 mg/mL at the first milking and between 0.8 and 46.1 mg/mL at the fifth milking. Distribution of milk yield and IgG concentration among quarters was not entirely repeatable during the first 5 successive milkings after parturition; that is, the ranking of quarters changed (intraclass correlation coefficients for quarter milk yield and IgG concentration: 0.64 and 0.79, respectively). The average hourly milk production increased in all quarters, ranging from 0.02 to 0.26 kg/h between the first 2 milkings up to 0.11 to 0.45 kg/h between the fourth and fifth milkings. First colostrum yield was not affected by any of the evaluated cow- and calf-related factors. Quarter colostrum IgG concentration was higher in cows with a higher previous lactation yield, whereas a lower colostrum IgG content was observed in cows with a longer gestation period and consequently heavier calves. In conclusion, milk yield and IgG concentration of individual quarters varied considerably, and their distribution among quarters within cows was moderately repeatable in consecutive milkings and changed partially over time. The decline of IgG concentration was independent of the concomitant increase in milk secretion, with changes occurring at different rates in individual quarters. Our results confirm the independence of the single mammary quarters at the onset of lactation despite an identical exposure to endocrine stimuli.

Suitability of somatic cell count, electrical conductivity, and lactate dehydrogenase activity in foremilk before versus after alveolar milk ejection for mastitis detection

Mastitis is responsible for substantial economic loss and significant animal welfare concerns for the dairy industry. Sensors that measure electrical conductivity (EC) and enzyme concentrations of lactate dehydrogenase (LDH) are presently used for automatic detection of mastitis. However, EC is not sensitive enough to detect mastitis, and the ability of LDH activity to identify mastitis caused by different pathogens is a potential option that needs to be investigated. This study was conducted to test the following hypotheses: (a) strict foremilk before milk ejection is more informative in detecting mastitis, in general, than foremilk removed after cows were stimulated for milk ejection; and (b) the value of LDH activity as a mastitis indicator depends on the type of pathogen associated with the infection. Milk samples (before afternoon milking) from 48 Holstein-Friesian cows at the University of Sydney's dairy farm (Camden, New South Wales, Australia) with EC > 7.5 mS/cm in any of the 4 quarters were collected over a period of 2 mo. Quarter milk samples (n = 343) from 48 cows were collected manually in the automatic milking rotary in 3 steps: foremilk before (strict foremilk) and after milk ejection, followed by an aseptic sample for bacteriological culture. The EC (mS), LDH (U/L), SCC (cells/mL), and milk protein and fat content (%) of foremilk in both sampling times were compared and used as predictors for gram-positive and gram-negative mastitis. Quarter (n = 515) observations from 44 cows were analyzed using a logistic mixed or linear mixed model, with cow and quarter nested within cow as random effects. Milk from both sampling times was also assessed by producing a receiver operating characteristic (ROC) curve and calculating the area under the curve (AUC) to determine ability to detect mastitis. Overall, EC and LDH were greater and milk protein (%) was lower in strict foremilk than in milk fractions obtained after milk ejection. Data from strict foremilk samples had slightly higher AUC values (0.98 to 0.99 vs. 0.97 to 0.98, respectively) than did the after-ejection milk samples. Although gram-negative coliform mastitis had significantly higher LDH activity than did gram-positive mastitis (6.19 vs. 5.34 log₁₀ U/L), the robustness of this result is questionable due to limited sample size. We concluded that milk samples taken before ejection can influence major mastitis indicators, suggesting that automatic milking system sensors could be modified to monitor milk before ejection for more efficient mastitis detection.

The Acceleration and Deceleration Profiles of U-18 Women's Basketball Players during Competitive Matches

The ability of a player to perform high-intensity actions can be linked to common requirements of team sports, and the ability to accelerate can be an important factor in successfully facing the opponent. The aim of this study was to determine the acceleration and deceleration profiles of U-18 women's basketball players during competitive matches. This study categorized accelerations and decelerations by playing position and quarter. Forty-eight U-18 female basketball players from the same Spanish league participated in this study. Each player was equipped with a WimuPro^TM inertial device. Accelerations/decelerations were recorded. The number of accelerations and decelerations, intensity category, and type were recorded. These variables varied between quarters (first quarter, second quarter, third quarter, and fourth quarter) and playing positions (Guard, Forward and Center). The shorter but more intense accelerations took place in the last quarter, due to the tight results of the matches. Besides, players in the Guard positions performed more accelerations and their intensity was greater than that of other positions. An acceleration profile was established for the quarters of a basketball game, and was shown to depend on the playing position, being different for Guards, Forwards and Centers in U-18 women's basketball players.