Recipient Organization
EMGENISYS INC
1000 LOUISIANA ST
HOUSTON,TX 770026008
Performing Department
(N/A)
Non Technical Summary
Livestock producers face growing pressure to increase efficiency, reduce waste, and improve sustainability in meat and milk production. One major challenge is the inability to predict the sex of a calf before it is born. Current tools for sex selection--such as sperm sorting, embryo biopsy, or genetic engineering--are often expensive, invasive, and not widely adopted due to their complexity or limited accuracy. This proposal seeks support for EmGenisys to develop the world's first non-invasive, real-time tool that uses artificial intelligence and machine learningto predict the sex of a bovine embryo based on a simple 30-second video recorded on a smartphone mounted to a microscope. This low-cost technology would allow beef and dairy producers to make smarter breeding decisions earlier in the production cycle, resulting in more targeted outcomes--like producing more heifers for milk or more bulls for beef--without altering or harming the embryo.To build this tool, EmGenisys will collect thousands of embryo videos in both laboratory and real-world settings. These videos will be paired with known sex outcomes, then used to train and test a computer model that can recognize subtle biological differences between male and female embryos. Once the model is accurate and reliable, it will be integrated into EmGenisys' existing online platform, which veterinarians and embryo transfer technicians already use to assess embryo health. By simply uploading a video, users will receive a prediction of an embryo's sex within seconds--no lab work, no invasive procedures. This research combines agriculture and cutting-edge technology to help producers make more informed decisions, reduce unnecessary costs, and optimize herd performance. The broader impact of this work includes increased food system efficiency, reduced waste in livestock production, and improved economic outcomes for beef and dairy producers, particularly in rural communities. This tool will make advanced reproductive technologies more accessible, practical, and profitable for livestock operations of all sizes.
Animal Health Component
35%
Research Effort Categories
Basic
55%
Applied
35%
Developmental
10%
Goals / Objectives
This proposal requests support for EmGenisys to develop the world's first non-invasive, real-time tool to predict bovine embryo sex using machine learning (ML) analysis of 30-second video recordings of pre-implantation embryos. Current sex-selection technologies--including sperm sorting, embryo biopsy, and genetic engineering--are invasive, expensive, and often inaccessible to most producers due to technical and regulatory limitations. EmGenisys aims to address these challenges by creating a safe, practical, and field-deployable alternative that is compatible with mobile devices and optimized for real-world use. The system is designed for veterinarians and embryo transfer (ET) practitioners to capture video data directly from microscopes using smartphones, eliminating the need for specialized equipment. The platform automatically processes video data to extract morphokinetic patterns--subtle changes in embryo movement and development--correlated with sex, enabling real-time prediction without disrupting the embryo. This tool will empower cattle producers to make earlier, more strategic decisions in the breeding season, improving meat and milk production outcomes, reducing input costs, and supporting herd-level sustainability. To achieve this, the project includes three primary objectives. Objective 1 focuses on data collection: we will collect, record, and label video data of at least 400 in vitro-produced embryos in a controlled laboratory setting, where embryo sex will be confirmed using PCR. In parallel, we will conduct a large-scale field study with Yellowstone Genetics, recording 3,500 commercially produced embryos and tracking fetal sex via ultrasound after transfer. This multi-site approach ensures our model is trained with diverse, high-quality data representing both academic and commercial conditions. Objective 2 involves training, validating, and testing ML models to predict embryo sex using our expanded dataset. We will use computer vision techniques and custom ML architectures built on PyTorch to identify consistent sex-related patterns in the video data. Prior work has demonstrated success in predicting embryo viability using similar methods, and this study will extend those findings to build a sex classification model with high accuracy and generalizability. Objective 3 focuses on integration: the developed models will be embedded within EmGenisys' existing AWS-hosted embryo assessment platform, which currently provides embryo viability scores to customers across the U.S. The updated system will allow users to upload embryo videos, receive health and sex assessments, manage embryo data, and export reports--all from a single web-based interface. The platform will also include standardization tools for video recording settings, ensuring consistent results regardless of device. By the end of this project, EmGenisys will deliver a first-in-class, commercially viable embryo sexing solution that is non-invasive, affordable, and scalable for both lab and field use, advancing reproductive technologies in line with USDA's Animal Production and Protection priorities.
Project Methods
MethodsThis project will use both basic (60%) and applied (40%) research to develop and validate a non-invasive tool for predicting bovine embryo sex using machine learning (ML) analysis of video data. Data will be collected through in vitro laboratory studies and large-scale field trials to train and test predictive models, which will then be integrated into EmGenisys' cloud-based platform for real-time embryo sex prediction.In the laboratory studies, bovine ovaries will be collected post-mortem and transported to research sites. Oocytes will be aspirated, fertilized, and cultured using standard IVF procedures. On days 6 and 7 of development, 30-second videos of morula and blastocyst-stage embryos will be captured using smartphone cameras mounted to stereo microscopes. Embryo sex will be confirmed using PCR. In the field study, 3,500 IVP embryos will be produced by commercial labs during routine embryo transfer, evaluated, and recorded before transfer into recipient heifers. Fetal sex will be confirmed by ultrasound between days 60-75 of gestation. All video and outcome data will be uploaded to a secure AWS-hosted database.ML models will be developed using PyTorch. Two main approaches will be used: (1) morphokinetic tracking through background subtraction and motion quantification, and (2) pattern recognition using convolutional neural networks (CNNs) and temporal models such as 3D CNNs or LSTMs. Datasets will be split into training, validation, and test sets. Model performance will be evaluated using metrics including accuracy, sensitivity, specificity, precision, recall, and area under the ROC curve (AUC). Statistical analyses (e.g., paired t-tests, chi-square tests) will compare results across collection sites, device types, and production environments.Validated models will be deployed through EmGenisys' EmVision 360 platform, allowing users (ET practitioners and veterinarians) to upload videos and receive sex predictions in real time. The software will include automated video preprocessing, object detection, and reporting tools, and will be accessible via web and mobile interfaces. A focus will be placed on optimizing performance across various devices while maintaining consistent prediction quality.Efforts to Expand and Deliver Scientific KnowledgeEfforts to disseminate and apply the project's findings will include:Presenting research at professional meetings (AETA, IETS, AABP)Hosting live demonstrations at livestock shows, junior nationals, and the World Dairy ExpoPublishing results in peer-reviewed journals such as TheriogenologyOffering webinars, on-farm trainings, and onboarding sessionsCreating instructional videos and educational content for online distributionSupporting student learning through graduate research involvement and internshipsThese efforts will help translate scientific knowledge into practice while fostering industry adoption and future talent development.Evaluation PlanProject success will be evaluated using scientific, commercial, and educational key performance indicators (KPIs).Scientific Evaluation:Model Accuracy: Measured against confirmed sex outcomes from PCR or ultrasoundPrecision and Recall: To assess classification confidence and minimize false resultsModel Robustness: Evaluated across labs, devices, and embryo production conditionsDataset Size: Number of labeled embryo videos collected for model training and testingCommercial and Adoption KPIs:Customer Adoption: Number of new and existing users using sex prediction in EmVision 360Embryos Processed: Total number of embryos run through the system with predictions generatedRevenue Impact: Increase in per-embryo revenue with the addition of sexing (estimated +$30/embryo)Market Growth: Projected 35% growth in ET/IVF adoption due to increased utility of the platformEducational and Outreach KPIs:Events and Trainings: Number of conferences, webinars, and producer demonstrations conductedStudent Engagement: Number of students and interns trained through EmGenisys or academic partnersPublication Output: Number of abstracts submitted and peer-reviewed papers publishedEvaluation methods will include automated software usage tracking, surveys of practitioners and producers, user feedback interviews, and longitudinal analysis of industry adoption trends. Progress will be reviewed quarterly, and key milestones (model accuracy benchmarks, user growth, revenue targets) will be tracked to ensure the project remains on schedule and delivers measurable outcomes.Overall, this project will produce new scientific insights, deploy a practical and scalable AI solution, and deliver real-world impact through increased adoption of reproductive technologies in livestock. The funding requested will enable EmGenisys to conduct rigorous evaluation, scale operations, and translate this innovation into a broadly used, economically impactful product.