An update of the predicted lean yield equation for the Destron PG-100 optical grading probe.

ABSTRACT

The objective was to update the equation used for prediction of pork carcass leanness with the Destron PG-100 optical grading probe. A recent cutout study (completed in 2020-2021) consisting of 337 pork carcasses was used for this research. An updated equation was generated using a calibration dataset (N = 188 carcasses) and prediction precision and prediction accuracy of the new equation was evaluated using a validation dataset (N = 149 carcasses). The updated equation was generated using forward stepwise multiple regression selection techniques in PROC REG of SAS, and the same parameters as the existing equation were used to fit the model. The updated Destron equation [89.16298 - (1.63023 × backfat thickness) - (0.42126 × muscle depth) + (0.01930 × backfat thickness2) + (0.00308 × muscle depth2) + (0.00369 × backfat thickness × muscle depth)] and the existing Destron equation [68.1863 - (0.7833 × backfat thickness) + (0.0689 × muscle depth) + (0.0080 × backfat thickness2) - (0.0002 × muscle depth2) + (0.0006 × backfat thickness × muscle depth)] were similar in their prediction precision for determination of carcass lean yield (LY), with the updated equation R2 = 0.75 and root mean square error (RMSE) = 1.97 and the existing equation R2 = 0.75 and RMSE = 1.94. However, when prediction accuracy was evaluated using the variance explained by predictive models based on cross-validation (VEcv) and Legates and McCabe's efficiency coefficient (E1), the updated equation (VEcv = 67.97%; E1 = 42.41%) was much more accurate compared with the existing equation (VEcv = -117.53%; E1 = -69.24%). Furthermore, when accuracy was evaluated by separating carcasses into 3% carcass LY groupings ranging from less than 50% LY to greater than 62% LY, the existing equation correctly estimated carcass LY 8.1% of the time, while the updated equation correctly estimated carcass LY 47.7% of the time. In an effort to further compare the abilities of the updated equation, comparisons were made with an advanced automated ultrasonic scanner (AutoFom III), which scans the entire carcass. The prediction precision of the AutoFom III was R2 = 0.83 and RMSE = 1.61, while the AutoFom III correctly estimated carcass LY 38.2% of the time and prediction accuracy calculations for the AutoFom III were VEcv = 44.37% and E1 = 21.34%). Overall, refinement of the Destron PG-100 predicted LY equation did not change prediction precision, but substantially improved prediction accuracy.

In the swine industry, optical grading probes are used to collect measurements at one location on the carcass (i.e., the grading site); this information is used to predict carcass leanness with an established multiple regression equation. For the Destron PG-100 optical probe, the equation currently used by the industry was established using the 1992 Canadian National Cutout Study. The objective of the current study was to update the equation by utilizing a recent cutout study consisting of 337 pork carcasses (N = 188 for the calibration dataset; N = 149 for the validation dataset). A multiple regression equation was generated with the calibration dataset using stepwise techniques and the same parameters as the existing equation were used to fit the model. When the validation dataset was tested for prediction precision, the existing Destron predicted lean yield (LY) equation (R2 = 0.75) and the updated Destron predicted LY equation (R2 = 0.75) were strikingly similar in their abilities to predict carcass LY. However, when prediction accuracy was evaluated by separating carcasses into six 3% LY groupings, the existing Destron predicted LY equation correctly estimated LY 8.1% of the time, while the updated Destron predicted LY equation correctly estimated LY 47.7% of the time.