Agreement and reliability of dairy cow udder morphometrics between traditional measurements and measurements from a 3-dimensional scanner

The morphology of the dairy cow mammary gland can indicate risk of disease and can change in response to various conditions. Collecting traditional measurements with a tape measure is time-consuming and can be subject to observer error. The objective of this cross-sectional study was to assess the agreement and reliability between traditionally collected measurements of the hind udder with measurements collected using a 3-dimensional (3D) scanner. We hypothesized there would be good reliability and agreement between methods. Adult cows in tiestalls at the Michigan State University dairy farm were used (n = 40). On 3 consecutive days, trained observers (n = 3) used a soft measuring tape to perform 3 measurements of the udder: a single vertical measurement from the rear udder attachment to the base of the quarter, and separate horizontal measurements of each quarter (left and right) from the medial cleft to the hind leg juncture. Marks were placed on the udders to guide traditionally collected measurements. A separate observer collected 3D scans (Structure Sensor, Occipital) and measured the same areas using 3-matic software (n = 120; Materialise, Belgium). Cows with mastitis (n = 3) and one incomplete scan were removed from the data, resulting in 37 cows with 999 and 333 traditional and 3D measurements, respectively. Intraclass correlation coefficients (ICC) among observers and between traditionally determined measurements and 3D measured were analyzed. Bland-Altman analysis was used to assess bias between traditional and 3D measurements. Linear regression was used to assess proportional bias between methods. The interobserver reliability was good to excellent for traditional measurements (ICC ≥0.88). Reliability between traditional and 3D measurements was good to excellent for all observers (ICC ≥0.87). Bland-Altman assessment found minimal bias between traditional and 3D measurements, except for 1 observer measurement of the right quarter. As compared with measurements from the 3D scanner, observers A and C consistently underestimated vertical measurements and overestimated the quarter measurements. Of results, 93% to 96% were within the upper and lower bounds of agreement. Observer C had positive proportional bias for vertical and right quarter measurements, indicating that as udder size increased, traditional measurements were overestimated as compared with 3D. These results indicate that the 3D scanner has a good to excellent reliability and adequate agreement with minimal bias in relation with the traditional measurements and may be used to collect hind udder measurements in healthy cows. These methods could be used to improve detection and assessment of conditions such as mastitis, though more research and validation studies are needed.