Source: UNIVERSITY OF TENNESSEE submitted to NRP
DEVELOPMENT, VALIDATION, AND EVALUATION OF COMPUTER IMAGING FOR TICK DETECTION ON CATTLE
Sponsoring Institution
National Institute of Food and Agriculture
Project Status
COMPLETE
Funding Source
Reporting Frequency
Annual
Accession No.
1028813
Grant No.
2022-70006-37984
Cumulative Award Amt.
$293,834.00
Proposal No.
2022-03428
Multistate No.
(N/A)
Project Start Date
Sep 1, 2022
Project End Date
Feb 28, 2025
Grant Year
2022
Program Code
[ARDP]- Applied Research and Development Program
Recipient Organization
UNIVERSITY OF TENNESSEE
2621 MORGAN CIR
KNOXVILLE,TN 37996-4540
Performing Department
Entomology & Plant Path - RES
Non Technical Summary
This is a two-year applied research project focused on enhancing agricultural biosecurity. Tick-borne diseases affect 80% of the world's cattle population and ongoing tick-pathogen biosecurity threats to the United States include Texas Cattle fever ticks causing Texas Cattle fever and bont ticks causing Heartwater in cattle. Asian longhorned ticks (ALTs) represent a new and emerging pest of companion animals, livestock, and wildlife. Since its discovery in the US in 2017, ALTs have been recovered from seventeen states and its feeding has been linked to sheep death, calf death, and cattle infection with T. orientalis Ikeda. Direct damage from ALT blood feeding includes significant economic losses in livestock including animal death and animal weight loss, milk loss, and direct damage to pelts / hides. Current surveillance methods for ticks on cattle are labor intensive as personnel must restrain an animal and then risk personal injury via physically touching and searching for ticks on animals making tick surveillance limited to the head and neck of an animal or not possible. Thus, the overarching problem is that while many exotic ticks and associated pathogens are poised to enter the United States, the current methods for detecting ticks on cattle (termed scratching) is impractical for producers, difficult for veterinarians, and primarily limited to the Texas Cattle fever tick eradication program. The proposed project addresses this severely neglected and critical need to help stakeholders use practical decision-making tools for producing sustainable and healthy cattle and limiting the movement of exotic and invasive tick species. Our long-term goal is to develop a data-driven IPM plan that integrates active and passive surveillance and prevents establishment of exotic and invasive ticks at a new site. There is a neglected and critical need to develop IPM strategies for ticks and their pathogens at livestock markets; thus, the objective of this proposal is to develop an automated method for detection of ticks on cattle at livestock markets. We will test the hypothesis that ticks can be automatically detected using computer vision and when combined with manual inspection, will help improve the overall efficiency of tick detection. We propose to develop, validate, and evaluate an automated method for tick detection to allow producers to quickly and safely check their livestock for ticks. Our proposed research is innovative and transformative because we are addressing a neglected point for tick dispersal in different emergent regions.Project results will enable producers to make informed decisions about treating cattle for ticks and will provide the necessary foundation for future sustainable veterinary pest management programs for livestock. Knowing cattle primarily move through livestock markets and that these markets do not have an IPM plan for managing ticks, our project is providing much-needed information and guidelines for preventing movement and establishment of exotic invasive ticks at livestock markets, using ALTs as the justification. After the automated tick detection method is developed, validated, and evaluated, it can be used in different areas, evaluated for adoption and practicality, and improved as needed for specific pest-pathogen complexes on different hosts in different regions. The proposed research is in contrast to traditional tick management approaches that involve on-animal studies, as we are working with stakeholders at the site of dispersal. Given the project's collaborative and integrated partnership of academic faculty and stakeholders, we are set to apply our research results in different regions with additional tick species and for future studies on cattle and other livestock.
Animal Health Component
25%
Research Effort Categories
Basic
50%
Applied
25%
Developmental
25%
Classification

Knowledge Area (KA)Subject of Investigation (SOI)Field of Science (FOS)Percent
3123120113025%
3123310111025%
2163120113025%
2163310111025%
Goals / Objectives
We will use a combination of computational and field-based approaches to achieve the following research objectives. We will first develop hardware and software of an automated camera system to monitor ticks on a beef herd known to have tick problems at the University of Tennessee, and then validate the developed system at livestock markets in Tennessee and North Carolina (sites where ticks were previously collected). Additional collaborators at the University of Georgia and USDA-ARS in Kerrville agreed to monitor their cattle for ticks and assess the automated method for tick detection (see letters of support). Our long-term goal from this project is to develop a data-driven IPM plan that integrates active and passive surveillance and prevents establishment of exotic and invasive ticks (e.g., H. longicornis) at a new site. There is a severely neglected and critical need to develop IPM strategies for ticks and their pathogens at livestock markets; thus, the overall objective of this proposal is to develop an automated method for detection of ticks on cattle at livestock markets. We will test the hypothesis that, like other machine learning and automated methods that we have developed (Gan et al. 2018, Gan, Lee, Alchanatis, et al. 2020, Gan, Lee, Peres, et al. 2020, Nasiri et al. 2020, Psota et al. 2021), ticks can be automatically detected using computer vision that, when combined with manual inspection, will help improve the overall efficiency of tick detection. We are a prepared team with the expertise, current pest threats, access to tick-infested cattle, and collaborations to test our hypotheses with the following three objectives.Objective #1. Develop an automated computer vision system that detects and counts ticks on cattle in order to develop a producer friendly strategy for monitoring ticks. Our working hypothesis, based on our preliminary data, is that a neural network will aid in the rapid detection of ticks on cattle. As we have developed a camera system for monitoring cattle's body conditions, we will upgrade this system by adding cameras, optimizing the placement of the cameras, and developing new machine learning algorithms for tick detection. The system will be deployed and developed for cattle at a tick-infested farm owned by the University of Tennessee.Objective #2. Validate the resulting computer vision system by comparing tick detection via traditional methods (scratching) to our developed system in order to use this method at detecting ticks quickly and accurately. Our working hypothesis, based on our preliminary data, is that tick detection with the network will be faster, safer, and equally (if not better) at detecting ticks on cattle by having a higher specificity and sensitivity than the golden standard (tick scratching). At the above tick-infested farm, the center hosts a bull development and evaluation program followed by a livestock sale, we will validate the computer vision system against the gold standard as cattle pass through chutes at these events.Objective #3. Evaluate the developed computer vision system for detecting ticks on cattle at various livestock markets in order to implement a rapid plan for preventing tick and pathogen movement at livestock markets. Our working hypothesis, based on our previous experiences and preliminary data, is that the computer vision system will quickly inform livestock markets if animals are infested with ticks and provide the data for the market to use an acaricide to prevent the spread of those ticks. Evaluating the computer vision program at several livestock markets with various degrees of tick infestation and in collaboration with the Cattle Fever Tick Eradication Program will assist with rapid tick detection and trace backing which provides the first steps for reducing the chances for exotic and invasive ticks (e.g., ALTs and Texas Cattle fever ticks) to spread to new farms.The proposed research is innovative and transformative because we are applying a state-of-the-art AI approach in a novel setting to address a severely neglected point for tick and pathogen dispersal in different emergent regions. Our expected outcomes include enhanced ability to detect and manage ticks as well as information about dispersal and establishment of exotic and invasive tick species. We anticipate these outcomes will have a positive impact by providing the US cattle industry and stakeholders who use livestock markets a mechanism for detecting and managing exotic and invasive ticks, thereby preventing widespread negative health and economic impacts caused by tick feeding.
Project Methods
The computer vision network will be developed and validated at the University of Tennessee Middle Tennessee Research and Education Center (MTREC). Information about each sampled animal and its associated lot will be recorded. All collections will have baseline information requested by USDA-APHIS. All data will be stored in a relational database and mapped in ArcMap 10.6.1. We will use a computer vision system mounted at the location of a water supply which is easily powered, requires minimal maintenance efforts, and can capture high-resolution images when animals are recorded. Already, each animal visits the water supply several times a day, so that it is guaranteed that each animal will be assessed every day. Building on what we initiated, we will develop a computer vision system consisting of several high-resolution cameras and deploy the system at a water supply for cattle.Development. At MTREC there is a bull evaluation facility which uses a C-LOCKTM SmartScale for each waterer. The SmartScale forces animals to approach the waterer from one direction and automatically logs the weight and radio frequency identification (RFID) tag of each animal. A solar panel kit, also from C-LOCKTM is used to provide power to the SmartScale. The computer vision system will be developed and installed on or close to the SmartScale. Because the animals can only approach the water from one side, cameras can be placed permanently to capture different body areas on cattle. In order to capture the head, neck, and tail areas, the initial placement of the cameras is high; however, various camera placement will be tested for optimal image quality. The camera system will include a Raspberry Pi 4 (8 GB RAM) and high-resolution surveillance cameras. The exact camera model will be determined during the development phase. A custom designed enclosure will be 3D printed to house and protect the camera system. Power will be supplied by the same solar panel kit as the SmartScale. To optimize power usage, an ultrasonic sensor will be used to trigger camera recording. A Python algorithm will be developed to use the ultrasonic sensor readings to trigger the cameras when an animal is at the waterer. The position and the distance threshold that triggers the cameras will be determined by experimentation. The cameras will be triggered to take multiple images, e.g., five images, each time an animal is at the waterer. Data will be stored in a 1TB flash drive. We expect each animal to visit the waterer at least twice a day. Assuming 30 animals on each pasture, the total daily recording for each pasture should be less than 150 images. Based on our preliminary results and calculations, the flash drive will last for more than 1.5 years at which point it will be replaced.The goal of the computer vision system is to perform automated detection of ticks. This is an object detection task that has been studied extensively. There are several object detection deep learning architectures; however, to achieve high detection accuracy for a specific task, a comprehensive training dataset is important. We propose to build a comprehensive dataset by incorporating a diverse set of annotated images (e.g., geo-graphic, breed, lighting, tick species, environmental conditions, etc.). The project team, including two post-doctoral scholars and an undergraduate student, will be trained by an expert to label the images. An initial deep-learning model will be trained for tick detection using the dataset collected previously. While more training data are collected and labeled, they will be added to the dataset to improve the model accuracy. The accuracy of the model will be evaluated based on mean average precision (mAP), an evaluation standard in the field of deep learning. The outcome will be a lightweight deep learning model that can accurately detect various tick species in the southeast region.Resulting images will be annotated and counted by three individuals, along with the computer vision system, in order to determine interrater and intrarater reliability. Both the individuals and the neural network will repeat the process three times with the same set of pictures. The total number of ticks counted by an individual rater and the system for one picture will be averaged amongst the repetitions. The coefficient variation (CV) will be calculated by dividing the standard deviation by the mean for every rater's picture. The CV will be calculated to determine a rater's dispersion around the mean for each picture counted. A lower CV means the rater had a more precise estimate every repetition. Mixed model analysis of variance will be used to assess interrater reliability, with the random effect of image and rater*image, while blocking on the image. To assess intrarater reliability, a no variable model in SAS 9.4, (Cary, NC) will be used with random effect of intercept. An interclass correlation (ICC) will then be calculated for each rater by dividing the estimate by the residual. ICC will be used to determine the correlation within a rater.Validation. To validate the resulting automated method, to maximize the efficiency of collections, and to protect the safety of both the investigator and the animal we will work with MTREC animals as they routinely enter chutes. The greater of 25% of the total lot size or ten animals will be sampled during each event, and to avoid reducing the efficiency of the husbandry practices of the producer and/or market. Animals that pose a threat to the safety of itself or the investigator will not be sampled. Procedures will be approved through the University of Tennessee's IACUC. The computer vision system will be installed at the end of chutes. As cattle are briefly stopped there, the computer vision system will first take the image(s) of the animals. Then animals will be scratched for ticks at common attachment sites along the ears, head, neck, tail, legs, belly, and underside of the tail for 5 minutes to minimize host stress. A second person will record animal information, where the investigator safely searched for ticks, where ticks originated from on the animal, the time it took to check the animals for ticks, and if any other incidences occurred. Ticks will be stored in ethanol until counted and identified to species and life stage.We will compare the number of ticks recorded by the automated system to the gold standard (number of ticks collected) for each animal. Additionally, analyses will include presence/absence variables about location of ticks, if animal could be checked, etc. We will also determine the amount of time it took for a person to scratch a tick compared to the vision system to capture and process the images. Separate generalized linear-mixed models will be conducted for each response variable. In the model, assessment method will be the fixed effects and other variables associated with the animal (e.g., breed) and collection (e.g., time, season) will be random effects.Evaluation. We will work with collaborators at livestock markets in North Carolina and Georgia and at the Cattle Fever Tick Eradication Program in Texas to evaluate the efficacy of the automated vision system and improve its efficiency to ensure it can be adopted and used at other markets. We will first develop a video and informational sheet for use of the vision system. Once instructions are developed, we will set up the vision systems at livestock markets for the spring, summer, and fall animal sales. Livestock markets will be visited weekly in year two by our team and we will modify the automated vision system as suggested by the market crews. For quality control as well as for tests of sensitivity and specificity, we will also compare the number of ticks collected from cattle to the number of ticks detected from cattle.

Progress 09/01/23 to 08/31/24

Outputs
Target Audience:Our target audience includes livestock producers (beef and dairy) and researchers (government, academic, and industry). We attempt to contact additional audiences such as students in the classroom, extension agents, livestock markets or auctions,consumers, and the scientific community. Demographics of these audiences have not been assessed. Changes/Problems:Unfortunatly, this project has had a number of changes, problems, and delays; most of which are circumstances. Deviation from research plans - Our research center became increasingly concerned about the potential of longhorned ticks and Theileria orientalis Ikeda infesting the property and potentially harming animals. As such, they increased their efforts to prevent and manage longhorned ticks through the increased use of acaricides. This is making finding these ticks extremely difficult since all animals are now treated with a preventative chemical. As such, we are continuing to seek partnerships with other researchers and veterinarians who may find animals infested with ticks. Because of the concern with Ikeda, cameras were not set up at the University of Tennessee Beef Heifer development program intake day and we were not permitted onto the premises to assess tick populations on animals. What opportunities for training and professional development has the project provided?A graduate student in Entomology and Plant Pathology with an Agricultural Economics background is leading the annotation portion of the project. She has learned more about computer science and coding (R and python) and is gaining a greater understanding of and appreciation for these techniques. Using this project as a part of her dissertation, this student applied to become a Foundation for Food and Agriculture Research Fellow (FFAR) and is a part of the 2023 cohort. Additionally, she is a recipient of the University of Tennessee Fall 2023 Student/Faculty Research Award ($5,000), which will allow her to conduct focus groups with relevant stakeholders to obtain input regarding how they would/could use the detection software created by this project and what they'd like to see improved/changed. This is an invaluable opportunity for this student to engage with stakeholders and educate them on the importance of early tick detection. A postdoctoral research associate in the Biosystems Engineering and Soil Science is leading the development of the neural network for tick detection. He has learned the status of the tick infection on cattle in the United States and how it is affecting cattle health and production. As an engineer, this project allowed him to expand his understanding of animal production and pest management in animal production. Upcoming Training Opportunity: To build, validate, and evaluate the dataset of images and introduce undergraduate students to veterinary entomology, Trout Fryxell wrote and submitted a USDA-REEU proposal with veterinary entomologists at different locations (Texas, Georgia, and Kansas). We were notified during the summer of 2024 that our project was awarded and students will begin internships in summer 2025. Additionally, we received word from USDA APHIS that we will be able to train stakeholders about tick inspections, both scratching and with the system, starting in January 2025. How have the results been disseminated to communities of interest?Results have been shared with academic scientists at meetings. When we attend wildlife meetings and producer stakeholder meetings, we also talk about the use of the developed computer vision technology to encourage participation and for stakeholders to provide images of tick-infested animals. For example, we discuss this project in 1-page reports introducing our ideas and the topic with the Tennessee Cattle Business magazine. We shared information about our project at the following meetings: January 2023. S1076 Multi-State Hatch Project, Orlando FL February 2023. Invited Department Seminar at Texas A&M, Kerrville, TX. April 2023. Invited presentation at Livestock Precision Farming hosted by University of Tennessee, Knoxville TN. June 2023. Developing a SMART surveillance platform for fly and tick detection in beef cattle. Livestock Insect Workers Conference, Fredericksburg, TX. (student presentation) September 2023. Provided slides to Texas A&M faculty to present to the Texas Animal Health Commission November 2023. Developing and validating a S.M.A.R.T. surveillance platform for fly and tick detection in beef cattle. Entomological Society of America, National Harbor, MD. (student presentation) December 2023. Sustainable Pest Management and Developing Data-driven Decisions for Livestock Pests, delivered to stakeholders at the Beef and Forage center Research and Recommendation Meeting, Knoxville TN January 2024. S1076 Multi-State Hatch Project, Las Cruces, NM June 2024. Detection and response to Haemaphysalis longicornis and Theileria orientalis Ikeda on a cow-calf farm in Tennessee. Livestock Insect Workers Conference, Melbourne FL. (student presentation) In the summer of 2024, we began to connect with international communities of interest. Specifically, in June 2024 we met with scientists from the International Livestock Research Institute attending the Beef Improvement Federation Symposium and discussed potential collaborations. These collaborations may lead to use of tick-infested images of cattle from Kenya which can also help with their phenotyping projects. Lastly, we continue to write about the project for the Tennessee Cattlemen's Association (https://www.tncattle.org/digitaltcb). These articles do not have a formal acknowledgments or funding section. We are using them to disseminate information to producers about surveillance and to recruit them to submit images of tick-infested animals. What do you plan to do during the next reporting period to accomplish the goals?Continue to gather images of tick-infested cattle, annotate them, and use them to help improve detection (accuracy and use). We plan to work with international collaborators to help improve the project. Due to the longhorned tick and Ikeda in Tennessee, producers are rightly concerned that they may have these vectors and pathogens on their property, so they are increasing their tick management plans, making it harder for us to detect ticks on cattle. To still find ways to visualize ticks, we will work with producers in southern Texas (September) and capture images of tick-infested cattle to improve the dataset. We have also reached out to investigators at the International Livestock Research Institute in Nairobi, Kenya to obtain additional tick-infestation images. We expect this connection to lead to a different set of images that will help train the models and increase overall accuracy and precision. Once the computer vision system is developed (MAP >80%), we will develop a website that allows stakeholders to upload images and determine if ticks are infesting the animal in the image (or not). We have already begun meetings with UTIA Office of Information and Technology to build this secure site. Ideally, producers will take and then upload images onto the platform, they will also provide date and address of the image to initiate the multistate and regional tick surveillance program the country needs. We received additional funding from the University of Tennessee Graduate School for the graduate student to conduct focus groups with producers to understand their willingness to adopt the platform and to identify areas for platform improvement.

Impacts
What was accomplished under these goals? Objective 1.We captured and requested images of tick-infested animals from producers and researchers. We attempted to obtain images of tick-infested cattle from our Middle Tennessee Research and Education Center, but few cattle had ticks and this was likely due to finding Theileria orientalis Ikeda on a nearby farm so farm personnel increased tick management protocols. Thus far, we have collected 708 tick-infested images which includes cattle, dogs, deer, horses, rodents, reptiles, birds,and rabbits. Three personnel have annotated images, but two have been primarily responsible for annotating.(Model 1) Using the first 76 of those annotated images, we developed an initial deep-learning model for tick detection; currently, the vision system can detect ticks with mean average precision of 20%. The initial evaluation of the results shows that the system can detect the presence of ticks when there are few or many ticks on an animal. However, it only counts the number of ticks accurately when there are few ticks present on an animal. The accuracy of counting the ticks decreases when many ticks are adjacent to each other.(Model 2) We ran a second model using 262 annotated images which increased the overall mean average precision to 38.9%. This second model performs better on large objects (e.g., engorged ticks that have taken a blood meal, female ticks) compared to small objects (e.g., flat ticks that have not taken a blood meal, male ticks).(Model 3) We ran the third model using 206 training images and the overall mean average precision increased to 44.15%. This third model is performing better on objects with few ticks but struggles with images that have many ticks.(Model 4) We ran the forth model using 264 training images and the overall mean average precision remained the same as our Model 3. This shows that a significantly larger number of training images are necessary to continue to improve the model accuracy.Although a significant improvement in counting the number of ticks is necessary, the vision system still shows some improvements compared to manual tick inspection as people can detect ticks when many are present but may fail to detect ticks when few are present (minimizing false negatives). In addition, it is clear that model accuracy is improved when the number of training images are increased, which shows the model's ability to learn quickly using training data. As more images are collected and annotated, it is expected that the model accuracy will continue improving quickly. Objective 2. Validate the resulting computer vision system by comparing tick detection via traditional methods (scratching). Initial validation was conducted by personnel at our university research farm. Personnel scratched their animals and found 11 ticks on 9 infested animals. All of the animals went through a chute with cameras mounted around the chute to potentially detect tick-infested animals. We are currently assessing those videos to see if we could see ticks and if so, if the computer vision system could detect them too.Additionally, Trout Fryxell plans to travel to southern Texas to further validate the system by scratching cattle that may be infested with cattle fever ticks (September 2024).As mentioned in the previous report, we added an experimental component that compares annotations by three humans (time/accuracy) to the computer results. Objective 3. Evaluate the developed computer vision system for detecting ticks on cattle at various livestock markets. As the development and validation of the computer vision system improves to a mean average precision of 80%, we will begin to evaluate the system in livestock markets. Unfortunately, we believe the discovery of H. longicornis and T. orientalis Ikeda on many Tennessee farms has many producers treating cattle with more acaricides and managing vegetation to prevent population spread and establishment. This is making it difficult for us to find tick-infested animals so we are having to look elsewhere for them to validate and evaluate our system. The goal here is to minimize the number of false positives (system indicates infested but animals are tick free) and false negatives (system indicated animals are not infested but are actually infested)

Publications

  • Type: Other Status: Published Year Published: 2024 Citation: K. Smith and R. Trout Fryxell. January 2024. New Year, New Buzz: dispelling myths about flies and ticks. Tennessee Cattle Business. Vol. 39. No. 1. Pp. 22-23. non-peer reviewed articles for the Tennessee Cattlemens Association (https://www.tncattle.org/digitaltcb). These articles do not have a formal acknowledgments or funding section. We are using them to disseminate information to producers about surveillance and to recruit them to submit images of tick-infested animals.
  • Type: Other Status: Published Year Published: 2024 Citation: K. Smith and R. Trout Fryxell. March 2024. Asian longhorned tick and Theileriosis Updated. Tennessee Cattle Business. Vol. 39. No. 3. Pp. 32-33. non-peer reviewed articles for the Tennessee Cattlemens Association (https://www.tncattle.org/digitaltcb). These articles do not have a formal acknowledgments or funding section. We are using them to disseminate information to producers about surveillance and to recruit them to submit images of tick-infested animals.
  • Type: Other Status: Awaiting Publication Year Published: 2024 Citation: K. Smith and R. Trout Fryxell. August 2024. Highlights from the Livestock Insect Workers Conference. Tennessee Cattle Business. Awaiting publication non-peer reviewed articles for the Tennessee Cattlemens Association (https://www.tncattle.org/digitaltcb). These articles do not have a formal acknowledgments or funding section. We are using them to disseminate information to producers about surveillance and to recruit them to submit images of tick-infested animals.
  • Type: Other Status: Awaiting Publication Year Published: 2024 Citation: Smith, K. V., R. A. Butler, V. Kobbekaduwa, J. Chandler, L. Strickland, and R. T. Trout Fryxell. 2024. Longhorned tick and Bovine Theileriosis. University of Tennessee Institute of Agriculture Factsheet, W XXX. Awaiting publication


Progress 09/01/22 to 08/31/23

Outputs
Target Audience:Our target audience includes livestock producers (beef and dairy) and researchers (government, academic, and industry). We attempt to contact additional audiences such as students in the classroom, extension agents, livestock markets or auctions,consumers, and the scientific community. Demographics of these audiences have not been assessed. Changes/Problems:Major Delay- The program was funded to begin August 2022 with the hiring of a postdoctoral scholar to lead the image collection and annotation; however, the scholar was not hired until October 2023. To keep the project moving forward, a PhD graduate student began work on the project and now leads the image collection and annotation efforts. Major Change- Drs. Deidre Harmon and Wes Watson left North Carolina State University for an industry position and retirement. We have shifted their responsibilities of evaluating the computer vision platform back to the University of Tennessee. Deviation from research schedule- We had expected to begin capturing images of ticks on cattle in spring 2023 at the Middle Tennessee Research and Education Center. This was slightly delayed to summer 2023 when cattle could be moved into the area. However, there were fewer ticks than expected making image annotation difficult. We then purchased GoPro Cameras for the center producers to use while working animals. We have also begun to crowdsource tick-infested images by asking academics and stakeholders for images of ticks on their animals. This change helped us get a variety of tick species from different environments and on different animals. What opportunities for training and professional development has the project provided?Professional Development: A graduate student in Entomology and Plant Pathology with an Agricultural Economics background is now a part of the project. She has learned more about computer science and coding (R and python) and is gaining a greater understanding of and appreciation for these techniques. Using this project as a part of her dissertation, this student applied to become a Foundation for Food and Agriculture Research Fellow (FFAR) and is a part of the 2023 cohort. More about this student and the FFAR program can be found here (https://storymaps.arcgis.com/stories/73c3a1fd42c94b07b23814c4c870824e). Training Opportunity: To build, validate, and evaluate the dataset of images and introduce undergraduate students to veterinary entomology, Trout Fryxell wrote and submitted a USDA-REEU proposal with veterinary entomologists at different locations (Texas, Georgia, and Kansas). We should be notified of this project in spring 2024. How have the results been disseminated to communities of interest?Results have been shared with academic scientists at meetings. When we attend wildlife meetings and producer stakeholder meetings, we also talk about the use of the developed computer vision technology to encourage participation and for stakeholders to provide images of tick-infested animals. For example, we discuss this project in 1-page reports introducing our ideas and the topic with the Tennessee Cattle Business magazine Trout Fryxell and the graduate student will also share results with beef producers at the annual Tennessee Beef and Forage Recommendation meeting in December. What do you plan to do during the next reporting period to accomplish the goals?Continue to gather images of tick-infested cattle, annotate them, and use them to help improve detection (accuracy and use). We will set up cameras at the University of Tennessee Beef Heifer development program intake day as animals are driven through a chute for evaluation purposes. Once the computer vision system is developed, we will develop a website that allows stakeholders to upload images and determine if ticks are infesting the animal in the image (or not). To upload images and use the platform, stakeholders will provide date and address of the image to initiate the multistate and regional tick surveillance program. With additional funding (submitted), we will conduct focus groups with producers to understand their willingness to adopt the platform and to identify areas for platform improvement.

Impacts
What was accomplished under these goals? Objective 1- We captured and requested images of tick-infested animals from producers and researchers. We attempted to obtain images of tick-infested cattle from our Middle Tennessee Research and Education Center, but few images had ticks. Thus far, we have collected 262 tick-infested images which includes cattle, dogs, deer, and rabbits. Four personnel have annotated images (83 by person 1, 152 by person 2, 33 by person 3, and 33 by person 4). (Model 1) Using 76 of those annotated images, we developed an initial deep-learning model for tick detection; currently, the vision system can detect ticks with an overall accuracy of 20%. The initial evaluation of the results shows that the system can detect the presence of ticks when there are few or many ticks on an animal. However, it only counts the number of ticks accurately when there are few ticks present on an animal. The accuracy of counting the ticks decreases when many ticks are adjacent to each other. (Model 2) We ran the second model using all of theannotated images which increased the overall mean average precision to 38.9%. This second model performs betteron large objects (e.g., engorged ticks that have taken a blood meal, female ticks) compared to small objects (e.g., flat ticks that have not taken a blood meal, male ticks). Although a significant improvement in counting the number of ticks is necessary, the vision system still shows some improvements comparing to manual tick inspection as people can detect ticks when many are present but may fail to detect ticks when few are present (minimizing false negatives). The low accuracy in counting the ticks from our initial models is a typical problem when there is a small training dataset. As we continue to gather and annotate tick-infested images, we are confident that the accuracy will increase significantly. Objective 2- Validate the resulting computer vision system by comparing tick detection via traditional methods (scratching).We plan to validate the computer vision in year 2 of the project. Additionally, we have added an experimental component that compares annotations by four humans (time/accuracy) to the computer results. Objective 3-Evaluate the developed computer vision system for detecting ticks on cattle at various livestock markets.We plan to evaluate the developed computer vision system in year 2 of the project.

Publications

  • Type: Other Status: Published Year Published: 2023 Citation: We have written five non-peer reviewed articles for the Tennessee Cattlemens Association (https://www.tncattle.org/digitaltcb). These articles do not have a formal acknowledgments or funding section. We are using them to disseminate information to producers about surveillance and to recruit them to submit images of tick-infested animals. (1) K. Smith and R. Trout Fryxell. March 2023. Developing a Monitoring Plan for Pests on Beef Cattle. Tennessee Cattle Business, Vol. 38. No. 3. Pp. 28-29. (2) K. Smith and R. Trout Fryxell. May 2023. You are Correct, there are more Ticks in more Places. Tennessee Cattle Business, Vol. 38. No. 5. Pp. 20. (3) K. Smith and R. Trout Fryxell. July 2023. Long-Term Investments with Sustainable Benefits for the Management of Livestock Pests. Tennessee Cattle Business, Vol. 38. No. 7. Pp. 23. (4) K. Smith and R. Trout Fryxell. September 2023. Cattle, manure, flies, oh my! Tennessee Cattle Business, Vol. 38. No. 9. Pp. 28-29. (5) K. Smith and R. Trout Fryxell. November 2023. What happens to flies and ticks during winter? Tennessee Cattle Business. Vol. 38. No. 11. Pp. 24.