Technical note: Mammary gland ultrasonography to evaluate mammary parenchymal composition in prepubertal heifers

Bovine mammary gland development studies are often terminal or involve invasive biopsy procedures. Therefore, noninvasive means of assessing mammary development should be considered as alternative methods in live animals. The objective was to test if mammary ultrasonography can be used as a noninvasive way to estimate mammary parenchyma (PAR) composition in prepubertal dairy heifers with different average daily body weight gains. In the 84 d preceding, the ultrasound exam heifers were maintained in 1 of 3 treatment groups. Individual heifers were fed a high gain (1 kg/d; n = 6), low gain (0.5 kg/d, n = 6), or maintenance (n = 6) treatment diet. To achieve desired body weight gains, heifers were fed differing amounts of the same silage-based diet. Mammary glands of 18 crossbred heifers Holstein:Gyr underwent a single mammary ultrasound exam immediately before heifer slaughter, which took place when heifers weighed 142.0 +/- 8.0 kg and were 200 d old. The 4 mammary glands of each heifer were evaluated using a real-time B-mode ultrasound machine equipped with a 6.5-MHz micro convex transducer. Digital images (8-bit) of glands were obtained and PAR was identified within gland. Average pixel values per unit of PAR area were determined for each gland and analyzed at the level of heifer. Pixel results were interpreted on the basis that lower average pixel values reflect PAR with relatively high amounts of protein as opposed to fat. To help validate that the pixel value within PAR is associated with composition of PAR, pixel findings were compared with histological [number of adipocytes in PAR (Nad) and epithelial area in PAR (Ep)] and biochemical [percent crude protein in PAR (%CP), percent ether extract in PAR (%EE), PAR weight (WPAR), and mammary fat pad weight (WFAT)] composition of PAR in these same heifers. Within PAR, %EE and WFAT were positively correlated with pixel values, whereas %CP, Ep, and Nad were negatively correlated. Parenchyma weight did not correlate with pixel values. Regression analyses (fixed effect log-pixel value; random effect treatment) were used to estimate Nad, Ep, %CP, %EE, WPAR, and WFAT. Sensitivity analysis of regression equations revealed that accuracy of tested equations ranged from 0.77 to 0.93 and precision ranged from 0.56 to 0.82. Concordance correlation coefficients of the equations ranged from 0.41 to 0.76. In conclusion, ultrasonography of PAR can accurately measure and predict PAR composition in prepubertal dairy heifers growing at various rates of gain.