Prediction of intramuscular fat by ultrasound in lean cattle

To study the accuracy and precision of intramuscular fat prediction by ultrasound (USIMF) in lean cattle, prediction models were developed based on 325 pure and crossbred beef and dual purpose bulls originating from I) commercial herds (n = 180) and II) a performance test station (n = 145). The bulls were scanned across the 12th thoracic vertebrae using a Pie 200 SLC scanner. Five images were collected per individual bull for later processing in the lab. After slaughter, a 2.5 cm cross-sectional sample located by the 13th thoracic vertebrae was used for chemical analysis to determine intramuscular fat % (CHIMF; Mean 1.25%, SD 0.6%) at the image site. The data were analysed sequentially involving 1) the Total Dataset, 2) Dataset I and 3) Dataset II. Prediction models were developed separately for each dataset, using stepwise regression procedures. The validation model R2, RMSE, overall mean bias, SEP, rP and rank correlations for CHIMF and USIMF based on the best prediction model (Dataset I) were 0.48, 0.46%, − 0.02%, 0.46%, 0.70 and 0.67, respectively. Further improvement in prediction accuracy can probably be achieved by increasing the size of datasets and the CHIMF variation. The results strongly indicate that ultrasound has potential for IMF prediction in lean cattle, although the value in indirect selection of breeding cattle for meat quality needs to be further evaluated.