Animal Science Papers and Research
Permanent URI for this collection
Browse
Recent Submissions
Item Screening and selection of a machine learning algorithm for development of a model to select cows for clinical examination using data from automated health monitoring technologies and other predictors of cow healthM. M. Perez; E. Cabrera; Y. You; Y. Wang; K. Weinberger; D. V. Nydam; J. O. Giordano (Journal of Dairy Science, 2025-09-26)The objective of this study was to create a framework for training and selecting machine learning algorithms (MLA) to classify cow health status daily using data from multiple automated health monitoring systems (AHMS), including wearable and nonwearable sensors, combined with nonsensor data of potential value for predicting cow health. The work presented in this manuscript is part of a series of studies aimed at identifying a single candidate algorithm that, upon extensive refinement and further development, could be deployed in a commercial dairy operation to identify cows potentially affected by health disorders for clinical examination. Data from AHMS and other cow features and performance data, including the clinical health status of cows, were collected in a prospective cohort study including Holstein cows (n = 1,252). Data from AHMS used for MLA training included rumination, eating, and physical activity measured in the neck (neck sensor), temperature and physical activity measured in the reticulorumen (bolus sensor), physical activity and resting measured in the leg (leg sensor), and milk yield, milk electrical conductivity, and milk components (parlor sensors). Other non-AHMS data used were temperature and humidity index, cow and calving event features, and current and previous lactation performance and management indicators. The dataset included 22,415 cow-day records with 49 features. The dataset was split into training and testing sets in an 80:20 ratio, resulting in 17,887 and 4,528 cow-day records, respectively. Data imputation and standardization were applied automatically or manually. A diverse set of nondeep learning (n = 26) MLA were trained and compared using the open-source automated ML (AutoML) tool Lazy Predict Classifier (LZP). Upon selection of the best-performing nondeep learning algorithms (i.e., XGBoost, AdaBoost, Nearest Centroid, and Bernoulli Naive Bayes) from the pool tested with LZP, classifier algorithms were compared with more complex deep learning algorithms (multilayer perceptron, recurrent neural networks, long short-term memory networks, and gated recurrent unit models) not included in LZP. All algorithms underwent training and evaluation before selection of a single best-performing algorithm, using several metrics of performance. Ensemble learning models, particularly XGBoost, achieved the best performance and balanced results with a sensitivity of 82.4% and a precision of 42.6% combined with a specificity of 86.4% and a negative predictive value of 97.6%. This model also had the highest F1-score (0.56) and area under the curve (84.4%). The XGBoost algorithm also demonstrated robustness in handling missing data. Our comprehensive approach to MLA screening and selection enabled informed decisions in selecting a suitable algorithm for identifying cows for clinical examination based on the daily prediction of health disorder occurrence. The combination of the AutoML tool LZP and manual refinement and testing of multiple MLA provided a robust framework for comparing multiple ML models. Ensemble classification learner algorithms such as XGBoost and Adaboost might outperform other deep learning and nondeep learning algorithms for classifying cow health daily using AHMS and other cow management and performance indicators.Item SemenSolver spreadsheet: Version 2.0Adamchick, J.; Briggs, K. R.; Nydam, D. V. (2025-09-03)A dairy farm's breeding and semen choices impact dairy herd inventory three years down the road. It is complex to consider the consequences, tradeoffs, and dynamics of those choices, and it is costly to wind up with too few replacements to keep the dairy herd full with the most productive cows. This tool is intended to compare the youngstock inventory and cash flow under varying scenarios of number of potential replacement animals produced and raised as a result of various semen strategies, and the eventual use of those available replacements. It is an expansion of the original SemenSolver tool that was developed as a tool to guide week-by-week semen choices, customized for a specific farm, to capitalize on crossbred beef markets while maintaining the replacement supply to support herd size and performance. Original SemenSolver spreadsheet tool: https://hdl.handle.net/1813/115529 Original SemenSolver fact sheet: https://hdl.handle.net/1813/115530Item Cornell University Ruminant Center - Research summary reportCornell Animal Science (2025-06-30)The Cornell University Ruminant Center (CURC), updated in 2013 to replace the former Teaching and Research Center, is a commercial scale dairy farm located 15 miles east of Cornell’s main campus. Home to 570 milking cows and housing for 400 heifers and calves, CURC includes modern free-stall barns, a tie-stall barn for individual cow research, a centralized feed center, substantial silage storage area, long-term manure storage and over 2,600 acres of cropland, pasture and woodlands.Item Supplemental Material - Delaying induction of ovulation and timed AI in a Double-Ovsynch protocol increased expression of estrus and altered first service reproductive outcomes of lactating dairy cowsLaplacette, A. L.; Rial, C.; Sitko, E.; Perez, M.M.; Tompkins, S.; Stangaferro, M.L.; Thomas, M.J.; Giordano, J.O. (Elsevier, 2024-10-14)The objective of this randomized controlled experiment was to evaluate the effect of delaying induction of ovulation and timed artificial insemination (TAI) on expression of estrus before AI and first service reproductive outcomes. A secondary objective was to evaluate the effects of delaying induction of ovulation in a Double-Ovsynch protocol on ovarian function. Lactating Holstein cows (n = 4,672) from 2 commercial dairy farms fitted with sensors for automated detection of estrus were synchronized with a Double-Ovsynch protocol up to the first PGF2α (PGF-L) of the Breeding-Ovsynch portion of the protocol (Pre-Ovsynch: GnRH, 7 d later PGF2α, 3 d later GnRH, 7 d later Breeding-Ovsynch: GnRH, 7 d later PGF2α, 1 d later PGF2α). At PGF-L, cows blocked by parity (primiparous vs. multiparous) and semen used for first service (sex-sorted dairy vs. conventional beef) were randomly assigned to the G56 (n = 2,338) or G80 (n = 2,334) treatments. Cows in G56 had 56 h whereas cows in G80 had 80 h from PGF-L to induction of ovulation with the last GnRH (GnRH2) before AI. For both treatments, TAI occurred ∼16 h after GnRH2. All cows with automated estrus alerts between PGF-L and TAI were inseminated at detected estrus (AIE) without GnRH. Ovarian function and responses to synchronization were monitored based on circulating concentrations of progesterone and examination of the ovaries by ultrasonography. Data for binary outcomes were analyzed by logistic and continuous outcomes with lineal regression. More cows in G80 received AIE and had estrus before AI. Overall, pregnancies per AI (P/AI) did not differ for the G80 and G56 treatments. Cows in G80 that received TAI and had no estrus had fewer P/AI than cows with estrus that received AIE or TAI in G80, and fewer P/AI than cows AIE and cows that received TAI and had or did not have estrus in the G56 treatment. No differences were observed between treatments or for cows with and without estrus for pregnancy loss. Unlike some minor differences between treatments for concentrations of progesterone at GnRH2, the most notable differences in ovarian function were for cows in both treatments with or without estrus that received TAI. Cows with estrus, were more likely to have follicles > 16 mm, had larger follicles before ovulation, and had a greater ovulation risk after AI. Likewise, within the G80 treatment only, cows with estrus that received AIE or TAI had larger follicles, were more likely to have complete luteal regression, had greater ovulation risk, were more likely to have a functional corpus luteum, and had more circulating progesterone after AI. We concluded that delaying induction of ovulation and TAI was effective for allowing more cows to express estrus before AI which had different ovarian function outcomes and greater P/AI than cows that did not express estrus. However, the greater P/AI of cows that expressed estrus was insufficient to compensate for the reduced P/AI of cows that did not express estrus, and thus increase overall P/AI compared with the treatment without delayed induction of ovulation. Detection of estrus before AI in cows undergoing synchronization of ovulation could help identify cows with different likelihoods of pregnancy after insemination.Item SemenSolver: A tool to support dairy herd semen strategy under varying replacement needsAdamchick, J.; Briggs, K. R.; Nydam, D. V. (2024-09-10)Improved reproductive technologies and management have enabled dairy herds to produce more female calves than they need. One emerging strategy is to select a subset of superior animals to produce the next generation of herd replacements while breeding the others for off-farm value (often to beef semen). It is complex to anticipate how today’s breeding choices will impact herd inventory three years from now and it is costly to wind up with too few replacements to keep the dairy herd full with the most productive cows. We created the SemenSolver spreadsheet as a tool to guide week-by-week semen choices, customized for a specific farm, to capitalize on crossbred beef markets while maintaining the replacement supply to support herd size and performance.Item SemenSolver spreadsheetAdamchick, J.; Briggs, K. R.; Nydam, D. V. (2024-09-10)Improved reproductive technologies and management have enabled dairy herds to produce more female calves than they need. One emerging strategy is to select a subset of superior animals to produce the next generation of herd replacements while breeding the others for off-farm value (often to beef semen). It is complex to anticipate how today’s breeding choices will impact herd inventory three years from now and it is costly to wind up with too few replacements to keep the dairy herd full with the most productive cows. We created the SemenSolver spreadsheet as a tool to guide week-by-week semen choices, customized for a specific farm, to capitalize on crossbred beef markets while maintaining the replacement supply to support herd size and performance. Factsheet can be found here.Item A randomized controlled trial of the effect of automated health monitoring based on rumination, activity, and milk yield alerts versus visual observation on herd health monitoring and performance outcomesC. Riala, M. L. Stangaferrob, M. J. Thomasb, and J. O. Giordanoa (Elsevier, 2024-09-24)A primary objective of this randomized trial was to compare the percentage of cows that underwent clinical examination and were diagnosed with clinical health disorders (CHD) with a health monitoring program that relied only on automated monitoring system alerts vs a program that relied only on visual observation of clinical signs of disease to select cows for clinical examination. Another objective was to compare the effects of these health monitoring programs on milk yield, the herd exit dynamics (i.e., cows sold and dead), and first service reproductive outcomes. Lactating Holstein cows (n = 1,204) enrolled in the experiment were fitted with a neck-attached sensor of an automated monitoring system (HR Tags; Merck & Co., Inc) that generated health alerts based on rumination time and activity. Milk yield was monitored three times per day by automated milk meters (MM27BC, DeLaval). Cows were blocked by parity, close-up period diet, and stratified by previous lactation milk yield, and then were randomly assigned within block to different programs for monitoring health from 3 to 21 d in milk (DIM). Cows in the visual observation group (VO; n = 597) were selected for clinical examination exclusively based on visual observation of clinical signs of disease, whereas cows in the automated health monitoring group (AHM; n = 607) were selected for clinical examination based on health alerts consisting of the following: a Health Index Score <86 arbitrary units, daily rumination <250 min, or a reduction of >20% in daily milk yield. Once selected for examination, the clinical exam was the same for both treatment groups. Binary data such as the occurrence of CHD, herd exit, and pregnancies per AI were analyzed with logistic regression. Daily and weekly milk yield were analyzed using ANOVA with repeated measurements. More cows underwent a clinical examination, more cows were diagnosed with at least one CHD, and more cows received treatment in the AHM than the VO treatment group. Cows in the AHM treatment had more accumulated milk than cows in the VO treatment from 2 to 21 DIM. Cows in the AHM treatment diagnosed with at least one CHD produced more milk from 3 to 18 and 20 to 21 DIM than cows diagnosed with a CHD in the VO treatment. Fewer cows left the herd up to 21 DIM for the AHM than the VO treatment. Pregnancies per AI at first service were greater for the VO than the AHM treatment at 30 d but not at 50 d after AI and no difference in pregnancy loss was detected. In conclusion, a health monitoring strategy that used automated health alerts increased the risk of undergoing clinical examination and having CHD diagnosed compared with a program that selected cows for clinical examination based exclusively on visual observation. Cows monitored with the program that relied on automated alerts also had greater milk yield in the first 21 DIM. Thus, monitoring cow health based on automated behavior and milk yield alerts might be a more effective alternative for health monitoring than exclusive use of visual observation of clinical signs of disease.Item Comparison of Social Behavior and Housing Condition Effects on Sociability Scores in LEWES and NY3 Mouse Lines Using a Three-Chamber Paradigm TestBayrakdarian, Sylvia (2023-05)Studying social behavior in mice is a crucial area of research in neuroscience, providing insights into information transmission between conspecifics and modulators of their behavior in addition to identifying what these social signals may control. However, little research has been done to compare social behaviors amongst strains of wild-derived mice such as NY3 and LEWES, which have been bred from mice caught in the wild compared to the typically utilized laboratory strains bred for generations within the laboratory. The objective of this research project is to differentiate between social behaviors, as measured by sociability scores, in these mouse lines using a three-chamber paradigm test (3CT). The study also aims to assess the effects of housing conditions, specifically between single-housing and pair-housing, on sociability given the implications of isolation as a social stressor, and thus making it important to understand the impact of housing conditions on social behavior in these mouse lines. The three-chamber paradigm test is a commonly used technique used for studying social behavior in mice which involves placing a test mouse in a chamber with three compartments and giving it the opportunity to interact with either a stranger mouse, the social stimulus, or an inanimate object, the non-social stimulus. The subsequently derived sociability score is a measure of how much time the test mouse spends in the compartment with the social stimulus compared to the time spent with the non-social stimulus, and is ultimately used as a measure of social behavior in mice. The outcomes of this study may be of significant value in future behavioral studies involving these or genetically similar mouse strains, as they may provide insight into the determinants of sociability and the ways in which housing conditions may affect such social behaviors.Item Supplementary figures for Reproductive physiological outcomes of dairy cows with different genomic merit for fertility: biomarkers, uterine health, endocrine status, estrus features, and response to ovarian synchronizationSitko, Emily; Laplacette, Ana; Duhatschek, Douglas; Rial, Clara; Perez, Martin M.; Tompkins, Sheridan; Kerwin, Allison L.; Giordano, J.O. (Journal Dairy Science, 2024-06-07)Our overarching objective was to characterize associations between genomic merit for fertility and the reproductive function of lactating dairy cows in a prospective cohort study. In this manuscript, we present results of the association between genomic merit for fertility and indicators of metabolic status and inflammation, uterine health, endocrine status, response to synchronization, and estrous behavior in dairy cows. Lactating Holstein cows entering their first (n = 82) or second (n = 37) lactation were enrolled at parturition and fitted with an ear-attached sensor for automated detection of estrus. Ear-notch tissue samples were collected from all cows and submitted for genotyping using a commercial genomic test. Based on genomic predicted transmitting ability values for daughter pregnancy rate (gDPR) cows were classified into a high (Hi-Fert; gDPR >0.6; n = 36), medium (Med-Fert; gDPR -1.3 to 0.6; n = 45), and low (Lo-Fert; gDPR <-1.3; n = 38) group. At 33 to 39 d in milk (DIM), cohorts of cows were enrolled in the Presynch-Ovsynch protocol for synchronization of estrus and ovulation. Body weights, body condition scores (BCS), and uterine health measurements (i.e., vaginal discharge, uterine cytology) were collected from parturition to 60 DIM and milk yield was collected through 90 DIM. Blood samples were collected weekly through 3 wk of lactation for analysis of β-hydroxybutyrate, non-esterified fatty acids, and haptoglobin plasma concentrations. Body weight, BCS, NEFA, BHB, and Haptoglobin were not associated with fertility groups from 1 to 9 wk after parturition. The proportion of cows classified as having endometritis at 33 to 36 DIM tended to be greater for the Lo-Fert than the Hi-Fert group. The proportion of cows that resumed cyclicity did not differ at any timepoint evaluated and there were no significant associations between probability or duration and intensity of estrus with fertility group. Cows of superior genetic merit for fertility were more likely to ovulate, have a functional CL, have greater circulating P4, and have larger ovulatory size than cows of inferior fertility potential at key time points during synchronization of estrus and ovulation. Despite observing numerical differences with potential performance consequences for the proportion of cows that responded to synchronization of ovulation and were both cyclic and responded to the Ovsynch portion of the synchronization protocol, we did not observe significant differences between fertility groups. Although not consistent and modest in magnitude, the collective physiological and endocrine differences observed suggested that cows of superior genetic fertility potential might have improved reproductive performance, at least in part, because of modestly improved endocrine status, uterine health, and ability to ovulateItem The ovarian function and endocrine phenotypes of lactating dairy cows during the estrous cycle were associated with genomic-enhanced predictions of fertility potentialSitko, Emily; Laplacette, Ana; Rial, Clara; Duhatschek, Douglas; Giordano, J.O.; Perez, Martin M.; Tompkins, Sheridan; Kerwin, Allison L.; Wiltbank, Milo C.; Domingues, Rafael R. (Journal of Dairy Science, 2024-04-29)The objectives of this prospective cohort study were to characterize associations among genomic merit for fertility with ovarian and endocrine function and the estrous behavior of dairy cows during an entire, non-hormonally manipulated estrous cycle. Lactating Holstein cows entering their first (n = 82) or second (n = 37) lactation had ear-notch tissue samples collected for genotyping using a commercial genomic test. Based on genomic predicted transmitting ability values for daughter pregnancy rate (gDPR) cows were classified into a high (Hi-Fert; gDPR >0.6 n = 36), medium (Med-Fert; gDPR -1.3 to 0.6 n = 45), and low fertility (Lo-Fert; gDPR <-1.3 n = 38) group. At 33 to 39 DIM, cohorts of cows were enrolled in the Presynch-Ovsynch protocol for synchronization of ovulation and initiation of a new estrous cycle. Thereafter, the ovarian function and endocrine dynamics were monitored daily until the next ovulation by transrectal ultrasonography and concentrations of progesterone (P4), estradiol, and FSH. Estrous behavior was monitored with an ear-attached automated estrus detection system that recorded physical activity and rumination time. Overall, we observed an association between fertility group and the ovarian and hormonal phenotype of dairy cows during the estrous cycle. Cows in the Hi-Fert group had greater circulating concentrations of P4 than cows in the Lo-Fert group from day 4 to 13 after induction of ovulation and from day -3 to -1 before the onset of luteolysis. The frequency of atypical estrous cycles was 3-fold greater for cows in the Lo-Fert than the Hi-Fert group. We also observed other modest associations between genomic merit for fertility with the follicular dynamics and estrous behavior. There were several associations between milk yield and parity with ovarian, endocrine, and estrous behavior phenotypes as cows with greater milk yield and in the second lactation were more likely to have unfavorable phenotypes. These results demonstrate that differences in reproductive performance between cows of different genomic merit for fertility classified based on gDPR may be partially associated with circulating concentrations of P4, the incidence of atypical phenotypes during the estrous cycles, and to a lesser extent the follicular wave dynamics. The observed physiological and endocrine phenotypes might help explain part of the differences in reproductive performance between cows of superior and inferior genomic merit for fertility.
- «
- 1 (current)
- 2
- 3
- »