Source: UNIV OF WISCONSIN submitted to NRP
COMBINING HIGH-THROUGHPUT PHENOTYPING AND GENOMIC INFORMATION OF CALVES TO IMPROVE PRODUCTIVE LIFE IN DAIRY CATTLE
Sponsoring Institution
National Institute of Food and Agriculture
Project Status
ACTIVE
Funding Source
AFRI COMPETITIVE GRANT
Reporting Frequency
Annual
Accession No.
1021996
Grant No.
2020-67015-30831
Cumulative Award Amt.
$500,000.00
Proposal No.
2019-06010
Multistate No.
(N/A)
Project Start Date
May 1, 2020
Project End Date
Apr 30, 2025
Grant Year
2020
Program Code
[A1231]- Animal Health and Production and Animal Products: Improved Nutritional Performance, Growth, and Lactation of Animals
Recipient Organization
UNIV OF WISCONSIN
21 N PARK ST STE 6401
MADISON,WI 53715-1218
Performing Department
Dairy Science
Non Technical Summary
Replacement heifers account for approximately 20% of the annual cost on dairy farms. As such, strategies to improve early selection of replacement heifers is crucial for the economic success of dairy operations. Currently, heifers are selected based on pedigree or genomic information. However, estimated genetic merit is an incomplete predictor of future performance because it does not account for other factors such as environment effects. The goal of this proposal is to provide a high-throughput phenotyping platform using computer vision systems to integrate phenotypes from biometric body measurements (BBM) and mammary gland (MG) development. Such phenotypes will be combined with genomic information to provide earlier and more accurate prediction of their future productive performance. The specific objectives are: 1) Further development and fine-tuning of an RFID-camera system for individual image-based body weight and BBM prediction for growth curve modeling of elite female replacement calves; 2) Development of an automated computer vision system to extract morphological characteristics of ultrasound images from mammary gland; 3) Combine BBM, assessment of MG development, and growth curve parameters, together with genomic information to predict reproductive and production performance of 1st lactation cows. The technology and knowledge generated from this project will allow a better monitoring of dairy calf development, providing an important tool for optimal management practices to improve decision-making regarding strategic heifer replacement and improve productive life performance of dairy cows. In addition, the computer vision system proposed herein could be used also as a tool for research development in other areas of dairy and animal sciences, such as physiology, nutrition, welfare, and health.
Animal Health Component
70%
Research Effort Categories
Basic
20%
Applied
70%
Developmental
10%
Classification

Knowledge Area (KA)Subject of Investigation (SOI)Field of Science (FOS)Percent
3023410101035%
3053410102035%
4027410208030%
Knowledge Area
302 - Nutrient Utilization in Animals; 305 - Animal Physiological Processes; 402 - Engineering Systems and Equipment;

Subject Of Investigation
3410 - Dairy cattle, live animal; 7410 - General technology;

Field Of Science
1010 - Nutrition and metabolism; 2080 - Mathematics and computer sciences; 1020 - Physiology;
Goals / Objectives
In this project, our main goal is to employ a high-throughput platform using computer vision systems to integrate phenotypes from biometric body measurements and mammary gland development. Such phenotypes will be used to assess real-time growth and development of dairy calves and, combined with genomic information, will provide more accurate prediction of their future productive performance. The following specific objectives are proposed in this project: (1) Further development and fine-tuning of an RFID-camera system for individual image-based body weight and biometric prediction for growth curve modeling of elite female replacement calves; (2) Development of an automated computer vision system to extract morphological characteristics of ultrasound images from the mammary gland; (3) Combine such biometric body measurements, assessment of mammary gland development, and growth curve parameters, together with genomic information to predict reproduction and production performance of 1st lactation cows.
Project Methods
Material and Methods from Objective 1: This first module of the project will be conducted at both Emmons Blaine Dairy Cattle Research Center (EBDCRC), in Arlington-WI, and Marshfield Agricultural Research Station (MARS), Stratford-WI. The system will be implemented in both facilities to collect daily image over the period of 4 weeks. Body weight and dorsal images will be collected from calves at EBDCRC through RFID-camera system. The total of 2,400 records of 600 animals of different ages will be used to update the predictive model. The calves will be weighed using a portable electronic scale. The images will be analyzed on the local server, and two strategies will be adopted regarding image segmentation. First, it will use the raw image (with no segmentation) and second, the image will be segmented. Two strategies will be adopted to implement feature extraction: 1) Biological Features, and 2) Computational features. The biological features here called biometric body measurements are those known to be associated with cattle BW, such as dorsal area, hip height, dorsal width, dorsal length and others. The computational features will be features extracted by deep learning algorithms, more specifically Variational Autoencoder. The features extracted (biological and computational) will be used as potential predictors of BW in different approaches: 1) Artificial Neural Networks (ANN); 2) Partial Least Squares (PLS); and 3) Multiple linear regression. After fine-tuning the models, an experiment will be conducted at the UW-Madison dairy facilities (EBDCRC and MARS) in order to implement the models and monitor growth development of dairy calves and heifers. For that experiment, we will recruit 200 female Holstein calves at the EBDCRC facility, which will be followed from birth to calving. Two nutritional plans will be randomly assigned for all calves, totalizing 100 calves per treatment. The nutritional plans will consist in high (HG) and low (LG) growth rates. Milk replacer will be reduced in both treatments to 50% at week 7, to induce weaning (week 8). Starter feed will be introduced at the end of week 4 and kept similar between treatments. After weaning, calves will be fed the same starter as fed in the pre-weaning period and water ad libitum. The main goal of the dietary treatment in this current research project is to create phenotypic variation on BW, biometric body measurements and mammary gland development. Health status will be assessed daily. All disease episodes and medical treatments will be recorded. Finally, BW prediction and biometric body measurements (e.g. dorsal area, height and width on 11 points along dorsal area, eccentricity (dorsal circularity), length, and others will be measured daily across time. Such predictions will be used to fit individual growth curves using alternative non-linear sigmoid curves.Material and Methods from Objective 2: We will collect weekly ultrasound images from mammary gland, from 1st to 8th week of life. At 2 months of age, mammary gland biopsies will be performed in all animals described in Objective 2 to characterize mammary gland cell proliferation and turnover. To accomplish that, gene expression of CASP3 (Caspase 3), CCND1 (Cyclin d), KRT8 (Cytokeratin 8), and PPIA (Cyclophilin A) will be measured in mammary gland tissue. After gene expression analyses, animals will be classified in high and low cell proliferation. The parenchyma tissue of all ultrasound images will be manually segmented and labelled based on the classification of high or low cell proliferation. In addition, labels regarding mammary gland parenchyma area and circularity will also be performed. Information regarding pixel brightness will be directly measured and collected as an important indicator phenotype, and for that variable predictions are not necessary. The segmented regions will be used as inputs to train a Convolutional Neural Network in order to predict cell proliferation classes, parenchyma area and circularity. The strategy of transfer learning will be used to train the CNN algorithm in this specific objective. Prediction quality will be assessed by Leave-One-Out Cross-Validation (LOOCV) within and across nutritional groups (HG and LG), and overall accuracy, precision, sensitivity (or recall), and specificity will be calculated.Material and Methods from Objective 3: The heifers' reproductive and productive performance is routinely recorded at the EBDCRC, including, days to 1st service, age at 1st service, days open, pregnancy at first and second insemination, pregnancy complications, gestation length, calving easy, milk yield and milk components. Early growth of calves will be monitored from birth to about 15 months of age and reproductive and production performance traits will be recorded from the breeding time until the end of the first lactation (about 34 months of age). Feed efficiency will be measured in all animals as Residual Feed Intake (RFI). In addition, all calves will be genotyped with the Zoetis ZL5 chip with 35,334 SNPs (Zoetis Services LLC, Parsippany - NJ) as part of the regular practices of the UW-Madison Arlington herd (EBDCRC), and GEBV predictions for each trait of interest will be obtained from the USDA-CDCB dairy genetic evaluation (https://www.cdcb.us/). All phenotypic traits collected in this research will be integrated in order to develop models to predict lactation performance and reproduction-related traits. Five set of variables will be tested regarding the capacity to accurately predict cows' lactation performance: Set 1: Use of mammary gland ultrasound images to predict total milk yield at first lactation; Set 2: Growth curve parameters; Set 3: Combining image-based features from mammary gland characteristics + body weight and biometric measurements during calfhood period + nutritional plan (calfhood period); Set 4: Set 3 + genomic information (GEBV); Set 5: Combining image-based features from mammary gland characteristicsnutritional plan (calfhood period) + growth curve parameters+ genomic information (GEBV).

Progress 05/01/23 to 04/30/24

Outputs
Target Audience:The target audiences during this period were researchers through scientific presentations. We also targeted dairy cattle producers, nutritionists, and veterinarians. The PI and Co-PIs of this project have actively participated in national and international conferences to present preliminary results of this project and to discuss the potential of animal phenotyping and genotyping to advance dairy cattle management. Changes/Problems:No changes or problems.A no-cost extension was requested and granted until 2024 to ensure that all experimental animals enrolled in the project complete a full first lactation period. What opportunities for training and professional development has the project provided?During this phase of the project, a PhD Student was trained on image analyses, genomic prediction, and machine learning. The student has been assisting on data collection and analyses as well as working on interpretation of results and writing manuscripts related to Aims 1, 2, and 3.The graduate student presented research abstracts at a scientific meeting, published one scientific paper, and is currently working on manuscripts. How have the results been disseminated to communities of interest?There has been significant dissemination of data to the research community during the 2023 reporting year. Abstracts were presented in person at the American Dairy Science Association meeting, US Precision Livestock Farming Conference (USPLF), American Society of Animal Science annual meeting, and many other conferences listed below. Additionally, two papers have been published, and two additional manuscripts are in progress for the 2024 reporting year. What do you plan to do during the next reporting period to accomplish the goals?We will continue to collect the data to complete an entire lactation of each individual animal.

Impacts
What was accomplished under these goals? In 2023 (the reporting year), all animals initiated their lactation period, and data on milk production, milk composition, and reproductive parameters were collected from all animals. During this period, we also evaluated mammary gland development during gestation and throughout lactation. Additionally, we developed algorithms for animal identification based on open-set systems. A no-cost extension was requested and granted until 2024 to ensure that all experimental animals enrolled in the project complete a full first lactation period. This research led to the publication of three scientific articles for the reporting year of 2023, which explored the application of ultrasound and histological imagery in understanding the mammary development of heifer calves.

Publications

  • Type: Journal Articles Status: Published Year Published: 2023 Citation: Vang, A.L., Bresolin, T., Frizzarini, W.S., Campolina, J.P., Menezes, G.L., Rosa, G.J., D�rea, J., & Hernandez, L.L. (2024). Monitoring mammary gland development in preweaned dairy heifers using ultrasound imaging. JDS Communications. https://doi.org/10.3168/jdsc.2024-0586
  • Type: Journal Articles Status: Published Year Published: 2023 Citation: Oliveira, D. A. B., T. Bresolin, S. G. Coelho, M. Magalhaes, C. Lage, L. G. R. Pereira, L. L. Hernandez, J. R. R. Dorea. 2023. A Polar Transformation Augmentation Approach for Enhancing Mammary Gland Segmentation in Ultrasound Images. Computer and Electronics in Agriculture. https://doi.org/10.1016/j.compag.2024.108825
  • Type: Journal Articles Status: Published Year Published: 2023 Citation: Menezes, G. L. Bresolin, T., Halfman, W., Sterry, R., Cauffman, A., Stuttgen, S., Schlesser, H., Nelson, M. A., Bjurstrom, A., Rosa, G. J. M., Dorea, J. R. R. Exploring associations among morphometric measurements, genetic group of sire, and performance of beef on dairy calves, Translational Animal Science, Volume 7, Issue 1, 2023, txad064, https://doi.org/10.1093/tas/txad064
  • Type: Conference Papers and Presentations Status: Published Year Published: 2023 Citation: Negreiro, A., R. E. P. Ferreira, T. Bresolin, G. J. M. Rosa, and J. R. R. Dorea. 2023. Computer vision system for identification of Holstein cattle during growth and across different physiological stages. Proceedings of the US Precision Livestock Farming, p.368. Knoxville-TN.
  • Type: Conference Papers and Presentations Status: Published Year Published: 2023 Citation: Negreiro, A., A. L. Vang, T. Bresolin, R. E. P. Ferreira, G. J. M. Rosa, L. L. Hernandez, and J. R. R. Dorea. 2023. Predicting age at puberty of dairy heifers based on biometric body features extracted from 3D images during the preweaning phase. J. Dairy Sci. 106:197. Annual Meeting ADSA, Ottawa, Canada.
  • Type: Conference Papers and Presentations Status: Published Year Published: 2023 Citation: Vang, A. L., W. S. Frizzarini, T. Bresolin, T. Cunha, G. L. Menezes, G. J. M. Rosa, L. L. Hernandez, and J. R. R. Dorea. 2023. Longitudinal histological and ultrasound analysis of bovine mammary gland development. J. Dairy Sci. 106:343. Annual Meeting ADSA, Ottawa, Canada.
  • Type: Conference Papers and Presentations Status: Published Year Published: 2023 Citation: Ferreira, R. E. P., and J. R. R. Dorea. Cloud computing to automate phenotype collection and data analyses in dairy systems. Proceedings of the US Precision Livestock Farming, p.131. Knoxville-TN.
  • Type: Conference Papers and Presentations Status: Published Year Published: 2023 Citation: Bresolin, T., R. E. P. Ferreira, G. J. M. Rosa, and J. R. R. Dorea. 2023. Computer vision on the edge: A computing framework for high-throughput phenotyping in livestock operations. Proceedings of the US Precision Livestock Farming, p.151. Knoxville-TN.
  • Type: Conference Papers and Presentations Status: Published Year Published: 2023 Citation: Menezes, G. L., A. Negreiro, R. E. P. Ferreira, and J. R. R. Dorea. 2023. Identifying dairy cows using body surface keypoints through supervised machine learning. Proceedings of the US Precision Livestock Farming, p.360. Knoxville-TN.
  • Type: Conference Papers and Presentations Status: Published Year Published: 2023 Citation: Ferreira, R. E. P., T. Bresolin, P. L. J. Monteiro, M. C. Wiltbank, and J. R. R. Dorea. 2023. Using computer vision to predict cyclicity of dairy cows during the transition period through 3D body surface images. J. Dairy Sci. 106:428. Annual Meeting ADSA, Ottawa, Canada.
  • Type: Conference Papers and Presentations Status: Published Year Published: 2023 Citation: Ferreira, R. E. P., M. C. Ferris, and J. R. R. Dorea. 2023. Optimizing training sets for individual identification of dairy cows. J. Dairy Sci. 106:429. Annual Meeting ADSA, Ottawa, Canada.
  • Type: Conference Papers and Presentations Status: Published Year Published: 2023 Citation: Artificial Intelligence for farm management and precision phenotyping. 46th ADSA Discover Conference: Milking the Data  Value Driven Dairy Farming. Chicago-IL, May 7th, 2024.
  • Type: Conference Papers and Presentations Status: Published Year Published: 2023 Citation: Transforming Dairy Farm Management with the Power of Artificial Intelligence. IDF World Dairy Summit. Chicago-IL, October 15th, 2023.
  • Type: Conference Papers and Presentations Status: Published Year Published: 2023 Citation: Digital technologies and Machine learning: A new way to look at novel traits at spatial and temporal dimensions. Annual Meeting of American Association of Animal Science (ASAS), Albuquerque-NM, July 22nd, 2023.
  • Type: Conference Papers and Presentations Status: Published Year Published: 2023 Citation: Precision Livestock Farming for Optimal Management. SIAM, Moroccan International Agricultural Show. Meknes, Morocco. May 4th, 2023.
  • Type: Conference Papers and Presentations Status: Published Year Published: 2023 Citation: Artificial Intelligence and Machine Learning for Agriculture. Summer School on Data Science. Fundacao Getulio Vargas. Rio de Janeiro, Brazil, Jan 24th, 2023.
  • Type: Conference Papers and Presentations Status: Published Year Published: 2023 Citation: Leveraging Artificial Intelligence to Optimize Farm Management Decisions. National Agricultural Producers Data Cooperative. University of Nebraska-Lincoln. September 19th, 2023.
  • Type: Conference Papers and Presentations Status: Published Year Published: 2023 Citation: Connectivity for Rural Areas and Sensing Technology Deployment. FCC Precision Ag Taskforce - Connectivity Demand. February 9th, 2023.
  • Type: Conference Papers and Presentations Status: Published Year Published: 2023 Citation: Computer vision and Machine Learning for Optimized Farm Management Decisions. CALS Data Science Showcase. November 1st, 2023.
  • Type: Conference Papers and Presentations Status: Published Year Published: 2023 Citation: AI from Foundations to Applications. Exploring Artificial Intelligence @ UWMadison. June 30th, 2023.
  • Type: Conference Papers and Presentations Status: Published Year Published: 2023 Citation: Shaping Futures with Data and Computing. Research Bazaar: Closing Panel. February 3rd, 2023.


Progress 05/01/22 to 04/30/23

Outputs
Target Audience:The target audiences during this period were researchers through scientific presentations. We also targeted dairy cattle producers, nutritionists, and veterinarians. The PI and Co-PIs of this project have actively participated in national and international conferences to present preliminary results of this project and to discuss the potential of animal phenotyping and genotyping to advance dairy cattle management. Changes/Problems:No changes or problems. What opportunities for training and professional development has the project provided?During this phase of the project, a Research Associate was trained on image analyses, genomic prediction, and machine learning, and a PhD student was hired. The Research Associate has been assisting on data collection and analyses as well as working on interpretation of results and writing manuscripts related to Aims 1, 2, and 3.The PhD student started working on Aims 1 and 2, and has been trained on computer vision systems, new laboratory techniques, and data analyses. The graduate student presented research abstracts at a scientific meeting, published one scientific paper, and is currently working on manuscripts. How have the results been disseminated to communities of interest?There has been significant dissemination of data to the research community during the 2022 reporting year. Abstracts were presented in person at the American Dairy Science Association meeting. Additionally, four papers have been published, and two additional manuscripts are in progress for the 2023 reporting year. What do you plan to do during the next reporting period to accomplish the goals?We will continue to work towards the goals.

Impacts
What was accomplished under these goals? For reference, during reporting year 2021: The algorithms to segment and extract animal body from 3D images were trained, which presented excellent accuracy. The 200 experimental animals were weaned, and the planned samples were collected (images, blood, genotypes, animal performance, health records, and mammary gland biopsies). The computer vision system composed by 30 depth cameras were deployed at UW research farm to perform the second phase of Aim 1, where body growth of heifers will be monitored through 3D images. All animals are being monitored and monthly body weights are being collected to re-assess the predictive performance of the computer vision algorithm. Throughout the year 2022, all 100 animals from the original pool of 200 calves were successfully inseminated and became pregnant. Our team initiated an extensive analysis of imaging data spanning from birth to puberty. This research led to the publication of a scientific article, which explored the application of ultrasound and histological imagery in understanding the mammary development of heifer calves. Furthermore, we conducted preliminary investigations to evaluate the relationship between body biometrics, captured via depth cameras within the first six weeks of life, and the age of puberty onset at 355 days, as well as the use of depth cameras to monitor body biometrics and weight. Comprehensive processing of all imaging data, in addition to biopsies and blood samples from birth to pregnancy, was diligently carried out in 2022.

Publications

  • Type: Journal Articles Status: Published Year Published: 2022 Citation: Vang, A. L., T. Bresolin, W. S. Frizzarini, G. L. Menezes, T. Cunha, G. J. M. Rosa, L. L. Hernandez, J. R. R. Dorea. Longitudinal analysis of bovine mammary gland development. Journal of Mammary Gland Biology and Neoplasia. 2023 May;28(1):11. DOI: 10.1007/s10911-023-09534-0.
  • Type: Journal Articles Status: Published Year Published: 2022 Citation: Ferreira, R. E. P., T. Bresolin, G. J. M. Rosa, J. R. R. Dorea. 2022. Using dorsal surface for individual identification of dairy calves through 3D deep learning algorithms. Computer and Electronics in Agriculture. 201:107272.
  • Type: Journal Articles Status: Published Year Published: 2022 Citation: Bresolin, T., R. E. P. Ferreira, F. Reyes, J. Van Os, J. R. R. Dorea. 2022. Feeding behavior of dairy heifers monitored through computer vision systems. Journal of Dairy Science. 106 (1), 664-675
  • Type: Journal Articles Status: Published Year Published: 2022 Citation: Caffarini, J. G., T. Bresolin, and J. R. R. Dorea. 2022. Predicting ribeye area and circularity in live calves through 3d image analyses of body surface. Journal of Animal Science, skac242. https://doi.org/10.1093/jas/skac24.
  • Type: Conference Papers and Presentations Status: Published Year Published: 2022 Citation: Dorea, J. R. R. Artificial intelligence for livestock systems. 2022. J. Dairy Sci. Vol. 105, Suppl. 1.
  • Type: Conference Papers and Presentations Status: Published Year Published: 2022 Citation: Bresolin, T., A. Wick-Lambert, R. E.P. Ferreira, A. Vang, D. Oliveira, G. Rosa, L. Hernandez, and J. R. R. Dorea. 2022. Phenotyping udder and mammary gland of dairy cows using computer vision systems. J. Dairy Sci. Vol. 105, Suppl. 1.
  • Type: Conference Papers and Presentations Status: Published Year Published: 2022 Citation: Silva, J. C. F., R. E. P. Ferreira, J. R. R. Dorea. 2022. Using computer vision for animal identification in dairy barns using isometric view images. J. Dairy Sci. Vol. 105, Suppl. 1.
  • Type: Conference Papers and Presentations Status: Published Year Published: 2022 Citation: Caffarini, J., T. Bresolin, and J. R. R. Dorea. 2022. Predicting ribeye area and shape of live calves through 3-dimensional image analyses of body surface. J. Dairy Sci. Vol. 105, Suppl. 1.
  • Type: Conference Papers and Presentations Status: Published Year Published: 2022 Citation: Negreiro, A., T. Bresolin, R. Ferreira, B. Dado-Senn, J. Laporta, J. Van Os, and J. R. R. Dorea. 2022. Monitoring heat stress behavior in dairy calves through computer vision systems. J. Dairy Sci. Vol. 105, Suppl. 1.
  • Type: Conference Papers and Presentations Status: Published Year Published: 2022 Citation: Dorea, J. R. R. and G. J. M. Rosa. 2022. Computer vision systems to advance high-throughput phenotyping in livestock. World Congress on Genetics Applied to Livestock Production. Rotterdam, Netherlands.
  • Type: Conference Papers and Presentations Status: Published Year Published: 2022 Citation: Bresolin, T., R. E. P. Ferreira. 2022. Effect of camera exposure time on image segmentation and body weight prediction. Journal of Animal Science https://doi.org/10.1093/jas/skac247.586
  • Type: Conference Papers and Presentations Status: Published Year Published: 2022 Citation: 2022 American Society of Animal Science - Midwest Section. March 15th, 2022 (50 attendees). Title: Artificial intelligence for livestock systems.
  • Type: Conference Papers and Presentations Status: Published Year Published: 2022 Citation: 2. Plant and Animal Genomics (PAG/ transferred to NRSP8: National Animal Genome Research Program/ Cattle and Swine). April 3rd, 2022 (100 attendees). Title: Challenges and opportunities of using computer vision systems for high-throughput phenotyping in dairy cattle.
  • Type: Conference Papers and Presentations Status: Published Year Published: 2022 Citation: REDTalk. School of Computer, Data & Information Sciences. April 21st, 2022 (60 attendees). Title: Harnessing the Power of Computer Vision System to Improve Management Decisions in Livestock Operations.
  • Type: Conference Papers and Presentations Status: Published Year Published: 2022 Citation: 2022 American Dairy Science Association annual meeting. June 24th, 2022 (150 attendees). Title: Artificial intelligence for livestock systems.
  • Type: Conference Papers and Presentations Status: Published Year Published: 2022 Citation: WCGALP12th World Congress on Genetics Applied to Livestock Production. July 4th, 2022 (150 attendees). Title: Use of High-Throughput Phenotyping and Data Analytics to Create Decision-Making Tools in Livestock Systems
  • Type: Conference Papers and Presentations Status: Published Year Published: 2022 Citation: ML+X Forum. American Family Insurance Data Science Institute. October 4th, 2022 (20 attendees). Title: Computer Vision and Machine Learning for Animal Farming.
  • Type: Conference Papers and Presentations Status: Published Year Published: 2022 Citation: Microsoft Research Summit. AI for Digital Agriculture. November 9th, 2022. (340 attendees). Title: Harnessing the Power of Computer Vision System to Improve Management Decisions in Livestock Operations.
  • Type: Conference Papers and Presentations Status: Published Year Published: 2022 Citation: Dairy Cattle Reproduction Council (DCRC). November 15th, 2022 (80 attendees). Title: Harnessing the Power of Computer Vision System to Improve Management Decisions in Livestock Operations
  • Type: Conference Papers and Presentations Status: Published Year Published: 2022 Citation: Fundacao Getulio Vargas. Summer Schoon on Data Science. Jan 24th, 2023. Title: Artificial intelligence for livestock systems.


Progress 05/01/21 to 04/30/22

Outputs
Target Audience:The target audiences during this period were researchers through scientific presentations. We also targeted dairy cattle producers, nutritionists, and veterinarians. The PI and Co-PIs of this project have actively participated in national and international conferences to present preliminary results of this project and to discuss the potential of animal phenotyping and genotyping to advance dairy cattle management. Changes/Problems: Nothing Reported What opportunities for training and professional development has the project provided?During this phase of the project, a research associate was trained on image analyses, genomic prediction, and machine learning, and a PhD student was hired. The research associate has been assisting on data collection and analyses as well as working on interpretation and writing manuscripts related to Aim 1 and 2.The PhD student started working on Aim 1 and 2, and has been trained on computer vision systems, new laboratory techniques, and data analyses. The graduate student presented research abstracts at a scientific meeting and is working on manuscripts. How have the results been disseminated to communities of interest?There has been significant dissemination of data to the research community during the 2021 reporting year. Abstracts were presented virtually (due to COVID19) at the American Dairy Science Association meeting. Additionally, one paper was published, and two more manuscripts are in progress. What do you plan to do during the next reporting period to accomplish the goals?We will continue to work towards the goals.

Impacts
What was accomplished under these goals? The algorithms to segment and extract animal body from 3D images were trained and presented excellent accuracy. The 200 experimental animals were weaned, and the planned samples were collected (images, blood, genotypes, animal performance, health records, and mammary gland biopsies). The computer vision system composed by 30 depth cameras were deployed at UW research farm to perform the second phase of Aim 1, where body growth of heifers will be monitored through 3D images. All animals are being monitored and monthly body weights are being collected to re-assess the predictive performance of the computer vision prediction.

Publications

  • Type: Journal Articles Status: Accepted Year Published: 2021 Citation: Oliveira, D. B. O., L. G. R. Pereira, T. Bresolin, R. E. P. Ferreira, J. R. R. Dorea. 2021. A Review of Deep Learning Algorithms for Computer Vision Systems in Livestock. Livestock Science, 253, 1:15.
  • Type: Conference Papers and Presentations Status: Accepted Year Published: 2021 Citation: Vang, A. L., T. Bresolin, W. Frizzarini, J. Campolina, G. J. M. Rosa, L. L. Hernandez, J. R. R. Dorea. 2021. Histological and ultrasound analysis of Holstein calf mammary gland development. AnnualJournal of Dairy Science 104: (Suppl. 1):218.
  • Type: Conference Papers and Presentations Status: Accepted Year Published: 2021 Citation: Oliveira, D. A. B, T. Bresolin, S. G. Coelho, M. Magalhaes, C. Lage, L. G. R. Pereira, L. Hernandez, and J. R. R. Dorea. 2021. Segmenting mammary gland tissue of preweaned dairy calves using spatial pyramid pooling networks. Journal of Dairy Science 104: (Suppl. 1):48.
  • Type: Conference Papers and Presentations Status: Accepted Year Published: 2021 Citation: Dorea, J. R. R., T. Bresolin, D. B. Oliveira, R. E. P. Ferreira. 2021. Harnessing the Power of Computer Vision System to Improve Management Decisions in Livestock Operations. Animal Science and Production Asociation, Padova-Italy.
  • Type: Conference Papers and Presentations Status: Accepted Year Published: 2021 Citation: Bresolin, T., F. Baier, J. Van Os, and J. R. R. Dorea. 2021. Feeding behavior of heifers monitored through computer vision systems. Journal of Dairy Science 104: (Suppl. 1):53.
  • Type: Conference Papers and Presentations Status: Accepted Year Published: 2021 Citation: Ferreira, R.E.P., T. Bresolin, L.G. Pereira, G. J. M. Rosa, and J. R. R. Dorea. 2020. Development of an identification system to recognize individual animals based on biometric facial features. Journal of Dairy Science 103: (Suppl. 1):240.
  • Type: Conference Papers and Presentations Status: Accepted Year Published: 2021 Citation: Ferreira, R.E.P., T. Bresolin, L.G. Pereira, and J. R. R. Dorea. 2020. Evaluating the predictive ability of point cloud deep learning to identify individual animals using surface-based body shape of dairy calves. Journal of Dairy Science 103: (Suppl. 1):52.
  • Type: Conference Papers and Presentations Status: Accepted Year Published: 2021 Citation: Dorea, J. R. R., T. Bresolin, R. E. P. Ferreira, L. G. R. Pereira. 2020. Harnessing the Power of Computer Vision System to Improve Management Decisions in Livestock Operations. Journal of Animal Science, 98: (Suppl. 4): 138-139.
  • Type: Conference Papers and Presentations Status: Accepted Year Published: 2021 Citation: Use of High-Throughput Phenotyping and Data Analytics to Create Decision-Making Tools in Livestock Systems. Animal Science and Production Association, University of Padova, September 21-24, Padova, Italy.
  • Type: Conference Papers and Presentations Status: Accepted Year Published: 2021 Citation: Enhancing livestock production systems through high-throughput phenotyping and data analytics. Annual Meeting of the Brazilian Society of Animal Science. August 10-14, Blumenau-SC, Brazil.
  • Type: Conference Papers and Presentations Status: Accepted Year Published: 2021 Citation: Image Analyses and Machine Learning in Livestock. HTCondor Week, University of Wisconsin-Madison, May 24-27. Madison-WI, USA
  • Type: Conference Papers and Presentations Status: Accepted Year Published: 2021 Citation: Harnessing the Power of Computer Vision System to Improve Management Decisions in Livestock Operations. Annual Meeting of American Association of Animal Science (ASAS), Madison, USA, July 2020.
  • Type: Conference Papers and Presentations Status: Accepted Year Published: 2021 Citation: Harnessing the Power of Computer Vision System to Improve Management Decisions in Livestock Operations. Michigan State University  Virtual Field Day on Precision Livestock Farming.
  • Type: Conference Papers and Presentations Status: Accepted Year Published: 2021 Citation: Challenges and opportunities of using computer vision systems for high-throughput phenotyping in dairy cattle. Iowa State University  Virtual Field Day on Genomic Selection and High-Throughput Phenotyping in Dairy Cattle. Agricultural Genome to Phenome Initiative (AG2PI).
  • Type: Conference Papers and Presentations Status: Accepted Year Published: 2021 Citation: Innovations in identifying glandular tissue. Institute for the Advancement of Breastfeeding and Lactation Education (IABLE), February 2022.


Progress 05/01/20 to 04/30/21

Outputs
Target Audience:During the reporting period, the PI (Dorea) and the co-PIs (Rosa and Hernandez) began reaching target audiences through research presentations. The PI and Co-PIs of this project reached agricultural professionals, scientists, and researchers working with high-throughput phenotyping, computer vision, genomic applications, lactation physiology, and farm management. We discussed the challenges and opportunities for high-throughput phenotyping for genetic selection and optimal farm management decisions in dairy cattle. Changes/Problems:The only major change is in the timeline. Due to the timing of when the award was initiated, conducting a study in summer 2020 was no longer possible. We planned to start in fall 2020, but due to the Covid-19 pandemic, we had to postpone the beginning of the study to summer 2021. Originally, we had proposed starting Aim 1 (with Modules 1 and 2) in Year 1 (summer 2020). Therefore, the PI team decided to conduct Module 1 of Aim 1 during fall 2020, and Module 2 of Aim 1 will start summer 2021, and it will be followed by the other aims. What opportunities for training and professional development has the project provided?The PI is directly supervising a Postdoctoral fellow and will recruit a PhD student to start in Summer 2021. In addition, an undergraduate student has been recruited for the Summer 2021 activities; she will assist with on-farm data collection and data processing. How have the results been disseminated to communities of interest? Nothing Reported What do you plan to do during the next reporting period to accomplish the goals?We will continue to work towards the goals.

Impacts
What was accomplished under these goals? (1) We collected more than 2,400 records of animal body weight and associated 3D, RGB, and infrared images (Module 1 - Aim 1). These images are being used to develop the models for body weight predictions that will be used in Module 2 (Aim 1). (2) We collected mammary gland ultrasound images from 30 calves under different diets to anticipate the development of deep learning algorithms to segment specific mammary gland tissues (e.g. parenchyma and mammary fat pad). This will optimize the use of human and computational resources, as we advance in the project timeline and collect ultrasound mammary gland images. (3) The IACUC protocol for the studies as well as the UW-Madison herd use request have been approved.

Publications