Work

Standardization of backfat thickness measurements across A-mode and B-mode ultrasound systems in pigs

Year: 2026

Type: article

Abstract: Backfat thickness is a key carcass trait used for economic valuation and genetic selection in pigs.However, field datasets are often generated using heterogeneous ultrasound devices, making direct comparison and integration difficult.This study evaluated the practical compatibility between an A-mode ultrasound device (PIGLOG 105) and a B-mode ultrasound device (ExaGo) for backfat thickness measurement and developed calibration/conversion models to standardize measurements to carcass reference scales.A total of 69 Yorkshire pigs from a domestic GGP farm were evaluated during July-August 2024.A-mode ultrasound measurements were collected 1-2 days prior to harvest at the 10-11th thoracic region.B-mode images were acquired at the identical location in two scan orientations: vertical (cross-sectional) and parallel (longitudinal).Images were stored in JPEG format and quantified using both Python-based analysis and ImageJ.Reference measurements included manual backfat thickness on hot carcasses (immediately post-harvest) and cold carcasses (after 24h chilling), VCS2000 outputs, and a direct measurement on the loin cut surface at the same anatomical level.Associations were assessed using Pearson correlation and simple linear regression with a zero intercept, and systematic bias was tested using paired t-tests.B-mode measurements showed closer agreement with reference values under vertical than parallel scanning.Slopes did not differ from unity for comparisons against the loin cut surface, which was measured at the same anatomical site, in three of the four B-mode combinations, whereas the smallest prediction error was observed for Python-vertical versus cold carcass.Despite high linearity, paired comparisons indicated significant systematic bias for VCS2000 and carcass references (p < 0.001), implying that raw B-mode values require calibration before pooling with reference-scale data.A-mode values were stably convertible to B-mode across four combinations (Python/ImageJ × vertical/parallel) using a single coefficient (β = 1.107-1.268;RMSE = 0.233-0.362).Direct A-mode prediction of reference measurements yielded slopes of 1.199-1.879,all significantly greater than unity (p < 0.001), and raw A-mode differed significantly from all references (mean differences 0.276-1.245cm;p < 0.001).After applying regressionbased calibration, mean differences were no longer significant (p = 0.142-0.528),leave-oneout cross-validation showed prediction errors only 0.6-1.1% larger than fitted errors, and concordance improved substantially (CCC 0.092-0.521to 0.483-0.745).In conclusion, Amode and B-mode are not directly interchangeable as raw data, but become compatible after applying the proposed conversion and calibration models, enabling practical standardization of field ultrasound measurements to carcass reference scales.

Source: Journal of Animal Science and Technology

DOI: 10.5187/jast.2600188