The benefits of CT scanning farm animals.
In the continuous drive to improve genetics and meat quality, medical imaging and diagnostic tools have long been adapted for agricultural research and application. CT scanning came to New Zealand in the 1990s when ultrasound scanning was already widespread.
In the 90’s agricultural scientist Dr Neville Jopson was at AgResearch when they were approached by Pāmu (then Landcorp NZ, prior to the split into Focus Genetics and Pāmu). Pāmu sheep operations are ‘vertically integrated’ – they breed genetically improved rams in their stud flocks for use in their commercial flocks and capture all the benefits of that genetic improvement when the commercial lambs are sent to the meat plant. They were using ultrasound to determine meat quality and yield in terminal sire stud flocks.
Ultrasound has been very beneficial for farmers and has progressed genetics significantly. It has a 40-50% accuracy predicting the weight of meat on a live animal and around 70% accuracy for fat content. Pāmu wanted better, and were looking to refine the odds, without having to rely on large slaughter trials. They wanted to explore the use of CT scanning. Serendipitously, Jopson had been working on the application of CT scanners for measuring livestock, and the InnerVision™ joint venture began along with installation of a CT scanner dedicated to agricultural science, at AgResearch Invermay in Mosgiel.
A CT scanner produces detailed cross-sectional images through an animal by rotating an X-ray tube through 360° and the scanner software reconstructs the two-dimensional image at that specific location. Images can be collected throughout the entire body which contributes to its high level of accuracy. CT images show muscle, fat and soft tissues, blood vessels, and bone. It is able to determine with accuracy the fat and lean in a carcass or meat cut and it does not require the slaughter of animals for the collection of the data.
A standard ultrasound is a blunter tool, used only at one measurement site on an animal, which limits its accuracy. However, ultrasound costs a few dollars per animal, compared to a few hundred dollars for CT scanning. Ultrasound is also a mobile operation – with CT scanning there’s no leaping in the truck and heading out to the farm, the farm needs to come to the CT scanner – bolted to the floor at AgResearch.
For Pāmu there’s been no sleep lost over the cost of scanning for terminal sires each season. They continue to use ultrasound on their farms, but as a diagnostic tool for genetic ranking to screen which rams should be CT scanned. Alongside genomic and weight datasets they select the top 10% of the young rams to truck into AgResearch for CT scanning. From the 200-250 animals they CT scan, the top 2% or so are identified to be used as sires in each stud. Jim Inglis of Pāmu says the economic benefit of CT scanning in their operation greatly outweighs the cost of scanning and hassle of trucking animals to the scanner.
In 2014 a new CT scanner was installed. This machine can image the entire sheep in a few minutes. The older machine was considerably slower (taking up to 2 hours to scan a whole animal) that they’d developed a 6 image set protocol rather than scanning the whole sheep. Now the ‘sticking point’ was the time it took for intensive analysis of the images by technicians — processing the CT images and calculating the weight of meat, fat and bone in the hindleg, loin and shoulder ‘primal cuts’.
Jopson wanted to reduce the time further, and with a couple of decades of CT scan images ‘in the bank’ he’d been trying to find the time to explore feeding images into a computer programme to ‘train’ a diagnostic application to read the scans. The COVID-19 lockdown was somewhat fortuitous for Jopson who finally had the time to sit down with an open source machine-learning platform, and 20 years of data.
Using the TensorFlow platform, Jopson has ‘trained’ an application to do the image analysis. He was surprised at how well, and how quickly, the programming worked, accurately analysing 95%+ of scans. It is also able to spit out data on the scans that could not be processed. Jopson can then analyse these, and end data can be returned to the application in order to further refine the end completion rate. The use of an open source platform means the application is not protected by IP, but the 20 years of analysed scans is — thanks to this ground breaking joint venture by AgResearch and Pāmu.
InnerVision™ provides the technology to other farms, vets, and researchers across New Zealand. CT scanning is being used in a variety of primary industry research beyond meat yields. For example, in one research programme, Pregnant ewes have been CT scanned to collate data on the effects of shearing during mid-pregnancy on a developing lamb.
The applications for research are exciting. A project is looking at methane production of individual sheep in order to aid genetic selection of low methane sheep. The CT scanner is able to image the rumen in situ and measure the volume of gases, liquid and churned up grass inside it. In the past it would have to be removed from the animal which involved both the loss of the animal and distortion of the rumen shape affecting accuracy of measurement.
Showdown Productions Ltd. Rural Delivery Series 16 2021