Automated sperm morphology analysis for vets enabled through machine learning

This project will enable data-driven decision making for farm vets, by automating the assessment of semen for morphological defects using Artificial Intelligence. This will improve the efficiency by which farm vets perform livestock fertility testing on farm, will improve livestock production rates, and standardises the approach for semen assessment. Currently, vets collect a semen sample during pre-breeding exams, prepare stained sample slides on-farm by using Eosin and spreading the semen sample onto a microscope slide. Then vets usually return to the laboratory with the stained slides, place them under a microscope equipped with high magnification objectives and then count the number of normal cells and abnormal cells on each sample. This takes around 5-10 minutes per sample, and requires significant expertise, preventing young vets to have confidence in their analysis and decisions. In the spring breeding season, UK vets working with beef farms may typically collect 30 samples a day meaning 2.5-5 hours of work for counting morphological defects upon their return to the laboratory.

In this project, Dyneval will build an AI platform for simple, fast, and accurate morphology semen assessments to be completed, allowing data-driven decisions to me made. We will use machine learning to develop deep neural network models that automatically classifies both normal cells and cells with morphological defects from microscopy images. This will save significant amount of time to the vets so that they can focus on other tasks for their business, and will allow data-driven decision to be made. Ultimately, we will aim to run the AI platform from images collected with Dyneval's innovative technology, the Dynescan which has been launched to the beef and dairy market in 2022 and uses a unique approach for assessing %motility and cell concentration, the other two key parameters considered within semen quality analysis.