Blab AI
Ohio State Researchers Develop AI Tool to Improve Cattle IVF Success
At Ohio State University, scientists in the Department of Food, Agricultural, and Biological Engineering have built a machine‑learning model that grades bovine eggs with steadier precision than human technicians, a breakthrough that could raise the success rate of in‑vitro fertilisation (IVF) in cattle.
IVF is a cornerstone for rapidly spreading elite genetics in beef and dairy herds. The process begins with the collection of eggs, or oocytes, from donor cows, fertilises them in a laboratory, and then transfers the resulting embryos into recipient females. While the technology can multiply desirable traits quickly, its overall effectiveness is limited by inconsistent embryo quality and lower conception rates compared with conventional breeding.
A major source of variability lies in the pre‑fertilisation grading of cumulus‑oocyte complexes (COCs). Technicians examine each egg under a microscope and assign a quality grade based on visual cues. The International Embryo Transfer Society (IETS) provides a grading scale, but the assessment is largely subjective and can differ between observers.
To address this, the FABE team partnered with Simplot, which supplied laboratory resources, personnel, and oocytes for analysis. The researchers compiled a dataset of more than 1,200 images of bovine oocytes that had been graded according to IETS standards. They then trained the YOLOv8 object‑detection algorithm, a state‑of‑the‑art deep‑learning model, to classify the images.
When the model was asked to reproduce the detailed IETS grades, it achieved about 63 % accuracy. When the grading categories were simplified into broader groups—“excellent,” “good,” and “bad”—accuracy rose to between 66 % and 80 %. Although these figures indicate room for improvement, they demonstrate that an AI system can provide a more objective and repeatable assessment than human graders.
Consistent grading is expected to improve downstream outcomes. By reliably identifying higher‑quality oocytes, the model could increase the proportion of embryos that develop normally and ultimately improve conception rates after transfer. The research team also envisions using machine‑learning models to predict the probability of embryo success for each oocyte, a capability that would help producers and laboratories make more informed decisions.
The researchers note that pairing AI grading with other technologies—such as time‑lapse embryo monitoring—could move the industry toward a more automated, precision‑driven IVF workflow. For cattle producers in Ohio and beyond, the technology could reduce the cost per pregnancy and accelerate genetic progress by making IVF a more dependable option.
The study, led by Professor John Fulton and former graduate student Grace Koppleman, represents a practical step toward improving reproductive efficiency in livestock. While the AI model is still in the research phase, the findings suggest that data‑driven decision‑making can reduce human variability in oocyte selection and potentially lift the performance ceiling of bovine IVF.
Future work will focus on refining the model’s accuracy, expanding the image dataset, and integrating predictive analytics for embryo viability. The team also plans to evaluate the system’s impact on actual conception rates in commercial settings. Until those studies are completed, the technology remains a promising but unproven tool for enhancing cattle reproductive technology.
IVF is a cornerstone for rapidly spreading elite genetics in beef and dairy herds. The process begins with the collection of eggs, or oocytes, from donor cows, fertilises them in a laboratory, and then transfers the resulting embryos into recipient females. While the technology can multiply desirable traits quickly, its overall effectiveness is limited by inconsistent embryo quality and lower conception rates compared with conventional breeding.
A major source of variability lies in the pre‑fertilisation grading of cumulus‑oocyte complexes (COCs). Technicians examine each egg under a microscope and assign a quality grade based on visual cues. The International Embryo Transfer Society (IETS) provides a grading scale, but the assessment is largely subjective and can differ between observers.
To address this, the FABE team partnered with Simplot, which supplied laboratory resources, personnel, and oocytes for analysis. The researchers compiled a dataset of more than 1,200 images of bovine oocytes that had been graded according to IETS standards. They then trained the YOLOv8 object‑detection algorithm, a state‑of‑the‑art deep‑learning model, to classify the images.
When the model was asked to reproduce the detailed IETS grades, it achieved about 63 % accuracy. When the grading categories were simplified into broader groups—“excellent,” “good,” and “bad”—accuracy rose to between 66 % and 80 %. Although these figures indicate room for improvement, they demonstrate that an AI system can provide a more objective and repeatable assessment than human graders.
Consistent grading is expected to improve downstream outcomes. By reliably identifying higher‑quality oocytes, the model could increase the proportion of embryos that develop normally and ultimately improve conception rates after transfer. The research team also envisions using machine‑learning models to predict the probability of embryo success for each oocyte, a capability that would help producers and laboratories make more informed decisions.
The researchers note that pairing AI grading with other technologies—such as time‑lapse embryo monitoring—could move the industry toward a more automated, precision‑driven IVF workflow. For cattle producers in Ohio and beyond, the technology could reduce the cost per pregnancy and accelerate genetic progress by making IVF a more dependable option.
The study, led by Professor John Fulton and former graduate student Grace Koppleman, represents a practical step toward improving reproductive efficiency in livestock. While the AI model is still in the research phase, the findings suggest that data‑driven decision‑making can reduce human variability in oocyte selection and potentially lift the performance ceiling of bovine IVF.
Future work will focus on refining the model’s accuracy, expanding the image dataset, and integrating predictive analytics for embryo viability. The team also plans to evaluate the system’s impact on actual conception rates in commercial settings. Until those studies are completed, the technology remains a promising but unproven tool for enhancing cattle reproductive technology.
