AgResearch shapes industry with cutting-edge genomics and imaging technologies

AgResearch animal genomics researchers are unlocking the potential of Artificial Intelligence (AI) to revolutionise the way we understand and optimise livestock production. AgResearch is using computed tomography (CT) scanning coupled with AI to achieve transformative advancements in our understanding of important animal traits that have previously been out of reach due to the complexity of their measurement. 

Traditionally, the focus of animal genomics has been on understanding the genetic variation underpinning livestock traits associated with enhanced productivity. However, recent developments in the analysis of data gleaned from CT scanning technology has opened new possibilities for researchers.  

By analysing imaging and precision measurements of the rumen - the largest compartment of the stomach in animals – researchers are gaining valuable insights into this complex organ and interrogate evidence that the weight and composition of rumen contents is associated with greenhouse gas emissions and feed efficiency.  

Because CT scanning is non-destructive, this allows animals to be measured, in some cases multiple times, and the animals are still available after measurement for breeding if required. Currently, many animals are already measured by CT scanning for carcass composition and meat quality traits, so the rumen information is also collected. 

Sam Hitchman, a Senior Data Scientist, used his background in medical imaging and data science to develop AI models for the AIU-driven automated segmentation using historical CT rumen data.  

This effectively means Sam has trained the latest AI models and tools to do a lot of the former leg work which we did manually (before AI). This has been made possible by the recently upgraded AgResearch e-research infrastructure which includes powerful Graphics Processing Units (GPUs). 

With AI-driven segmentation, Sam was able to overcome the limitations of manual methods and go back and use a lot of the previous collect data and extract new information from over 20 years of information gathering. 

“Our previous research was extremely thorough. It was a globally unique dataset. It included data that no other similar researchers had, due to lack of time and resource. So, when we customised DeepLabV3 (an off the shelf AI tool) and trained it to process our data we could provide our scientists, who were the first in the world to successfully develop low methane-emitting sheep genetic lines, with methods to rapidly screen their data in a much more accurate and detailed way. 

“For someone like me that is really rewarding. We can see the tangible and obvious way we are improving our research which isn’t always the case when our work is often so complex. The images, and how we analyse them, directly lead to better results down on the farm, and breeders understand that.” 

Beyond CT, AgResearch is exploring several other imaging technologies to improve animal productivity and welfare, including ultrasound and Hyperspectral imaging for meat quality, thermal cameras for disease detection, and more. AI processing and automation models can help bridge the gap between research trials and industry/farm deployment. It offers unparalleled efficiency and accuracy, freeing up valuable time for new research endeavours.  

Sam explains: “I don’t think it will be very long before farmers have thermal imaging cameras in their toolboxes. They will be able to mount them on stock yards and have an AI to identify, for example the cheek of a sheep, and from that thermal information and processing, they will be able to detect whether an animal is undergoing an animal health challenge. This is the type of research we are including in our research proposals. It has such a clear pathway to actually having impact on farm. It is just around the corner.” 

Stakeholders, including beef and lamb genetics companies, have been impressed by the capabilities of AI-driven segmentation and the impact of the technology. Moving forward, collaboration with key partners like Focus Genetics will be essential to drive adoption and maximize the impact of AI in animal genomics. Sam recently shared his research with the Scotland Rural College in Edinburgh who have a dedicated Computer Tomography Unit and was able to gain permission and access to use some of their data in his research. 

Sam eventually wants to make a complete CT model of a sheep carcass, including all its organs, which has recently been achieved in the medical field using extensive collections of human CT data. 

“I think that is the next natural progression. I can see huge benefits in having a model of, for example, a sheep’s liver or lungs, so that researchers can unpack and then make genetic gains to improve animal health for traits like facial eczema and pneumonia.” 

As AgResearch continues to push the boundaries of innovation in animal genomics, the importance of preprocessing data for AI cannot be overstated. By avoiding misleading patterns and adhering to the "garbage in, garbage out" principle, researchers can ensure the reliability and validity of their findings.  

The future of animal genomics research is further brightened, thanks to the transformative potential of AI. By leveraging cutting-edge technology and fostering collaboration, researchers are poised to unlock new insights and drive sustainable advancements in livestock production.