Murdoch University, School of Veterinary & Life Sciences, Western Australia 6150, Australia; Australian Cooperative Research Centre for Sheep Industry Innovation, Australia.
Meat and Livestock Australia, 40 Mount Street, North Sydney 2060, Australia.
Meat Sci. 2018 Oct;144:91-99. doi: 10.1016/j.meatsci.2018.06.020. Epub 2018 Jun 19.
This experiment assessed the ability of an on-line dual energy x-ray absorptiometer (DEXA) installed at a commercial abattoir to determine carcase composition at abattoir chain-speed. 607 lamb carcases from 7 slaughter groups were DEXA scanned and then scanned using computed tomography to determine the proportions of fat (CT fat%), lean (CT lean%), and bone (CT bone%). Data between slaughter groups were standardised relative to a synthetic phantom consisting of Nylon-6. Models were then trained within each dataset using hot carcase weight and DEXA value to predict CT composition, and then validated in the remaining datasets. Results from across-dataset validation tests demonstrated excellent precision for predicting CT fat%, with RMSE and R values of 1.32 and 0.89, compared to values of 1.69 and 0.69 for CT lean%, and 0.81 and 0.68 for CT bone% which had less precision. Accuracy across datasets was also robust, with average bias values of 0.66, 0.83, and 0.51 for CT fat%, lean%, and bone%.
本实验评估了安装在商业屠宰场的在线双能 X 射线吸收仪（DEXA）在屠宰链速下测定屠体成分的能力。对 7 个屠宰组的 607 个羊屠体进行了 DEXA 扫描，然后使用计算机断层扫描（CT）来确定脂肪（CT 脂肪%）、瘦肉（CT 瘦肉%）和骨骼（CT 骨骼%）的比例。将屠宰组之间的数据与由尼龙-6 组成的合成模型进行标准化。然后，在每个数据集内使用热屠体重量和 DEXA 值训练模型，以预测 CT 成分，然后在剩余的数据集内进行验证。跨数据集验证测试的结果表明，预测 CT 脂肪%的精度非常高，RMSE 和 R 值分别为 1.32 和 0.89，而 CT 瘦肉%的 RMSE 和 R 值分别为 1.69 和 0.69，CT 骨骼%的 RMSE 和 R 值分别为 0.81 和 0.68，精度较低。跨数据集的准确性也很稳健，CT 脂肪%、瘦肉%和骨骼%的平均偏差值分别为 0.66、0.83 和 0.51。
