Steve Connaughton

Dual Energy X-ray Absorptiometry (DXA) scanners operating at abattoir processing speeds are currently installed in six sheep meat abattoirs around Australia, predicting carcass composition as estimates of computed tomography (CT) determined fat %, lean %, and bone %. This study tested an updated bone-detection algorithm for these DXA scanners. This algorithm improved the precision of prediction for carcass fat% and lean%, but most notably for bone % (R2 = 0.92, RMSE = 0.61 %), compared to the previous algorithm (R2 = 0.51, RMSE = 1.57 %). This was due to improved allocation of bone-containing pixels, resulting from the inclusion of tissue thickness in the bone-detection equation. In a second experiment, the predictions from this new algorithm, along with an automated phantom calibration technique, were assessed relative to their ability to meet the AUS-MEAT accreditation accuracy standards required for predicting CT determined carcass fat%, lean%, and bone%. The DXA met these standards for predicting fat % (range 10.9 % - 37.1 %), lean % (range 49.0 % - 66.2 %), and bone % (range 11.6 % - 25.0 %), across three weight bands of light carcasses (<22 kg), mid-weight carcasses (22-28 kg), and heavy carcasses (>28 kg). This work allowed for the accreditation of DXA, enabling its predictions of carcass composition to be used for trading sheep carcasses in Australia. The accuracy of these predictions far exceed those provided by the historical industry measure of GR tissue depth, and hot carcass weight.

In 2016 an Australian project, the Advanced Livestock Measurement Technologies project (ALMTech), was initiated to accelerate the development and implementation of technologies that measure lean meat yield and eating quality. This led to the commercial testing, and implementation of a range of new technologies in the lamb, beef, and pork industries. For measuring lean meat yield %, these technologies included dual energy X-ray absorptiometry, hand-held microwave systems, and 3-D imaging systems. For measuring beef rib-eye traits and intramuscular fat %, both pre- and post-chilling technologies were developed. Post-chilling, a range of camera systems and near infrared spectrophotometers were developed. While pre-chilling, technologies included insertable needle probes, nuclear magnetic resonance, and X-ray systems. Initially these technologies were trained to predict the pre-existing traits already traded upon within industry. However, this approach was limiting because the technologies could measure attributes that were either non-existent in the trading language, were superior as calibrating standards, or more accurately reflected value than the pre-existing trait. Therefore, we introduced IMF% into the trading language for both beef and sheep meat, and carcase lean%, fat%, and bone% for sheep meat. These new technologies and the traits that they predict have delivered multiple benefits. Technology provider-companies are instilled with the confidence to commercialise due to the provision of achievable accreditation standards. Processors have the confidence to invest in these technologies and establish payment grids based upon their measurements. And lastly, it has enhanced data flow into genetic databases, industry data systems (MSA), and as feedback to producers.

A prototype, on-line Dual Energy X-ray Absorptiometer (DXA) has shown high precision of prediction of carcass composition for the purpose of improved sheep meat grading in the Australian lamb supply chain, albeit with small inaccuracies over time. These inaccuracies were present across hours, and more significantly across days, which were unacceptable for any accreditation of this device as an objective carcass measurement tool in Australia. This inaccuracy demanded the creation of a novel image processing algorithm for the prototype DXA. This DXA was tested for repeatability of predictions of lamb carcass composition over minutes, hours, and days, using two developed image processing algorithms. There was high immediate repeatability for both algorithms when predicting lean muscle % in 40 lamb carcasses, with a maximum coefficient of variation of 0.65% over five repeated scans. There was a decrease in the coefficient of variation of the prediction of lean muscle % of 30 lambs scanned three times over a 48-hour period from 5.93% to 1.19% when the superior algorithm was used. The inaccuracies of lean muscle % predictions were associated with increases in the unattenuated space pixel values in DXA images. Improvements of the current algorithm is required to demonstrate repeatability over time for the purpose of accreditation within the Australian sheep meat industry, and for possible expansion of this technology into international supply chains.

Dual energy x-ray absorptiometry (DXA) devices were installed at two Australian abattoirs to predict computed tomography (CT) determined fat % and lean % of lamb carcasses. This study tested three algorithms developed for these devices, termed β1, β2 and β3, and assessed their accuracy and precision in predicting CT composition. Algorithm β3 included the use of a plastic phantom calibration block scanned by both DXA devices to adjust prediction equations, resulting in superior accuracy to the algorithms without phantom calibration (β1 and β2). When compared to the gold-standard CT composition, the bias of the DXA predictions was lowest when using algorithm β3 at the two sites (-1.17%, -0.49% for fat %, 0.11%, -0.37% for lean %). The difference of DXA composition predictions between sites was lowest when using algorithm β3, which demonstrated between site differences of 0.59 CT fat %, and 0.46 CT lean%. In contrast, algorithm β1 and β2 produced differences of 23.7% and 30.8% for CT fat, and 17.3% and 21.9% for CT lean between the two DXA devices. There was a small difference of 0.78% between the fat predictions of the first DXA image compared to the second DXA image for each carcass. The precision of predictions improved slightly using algorithm β3. This work demonstrates that the in-line DXA systems can produce comparable results across sites.