Homogeneity density scores of quarter milk in automatic milking systems

Abstract

Milk quality and clinical mastitis in dairy cows are monitored by detecting visually abnormal milk. A standardized method to evaluate clots in milk and studies of the incidence and dynamics of clots in milk at the quarter level are lacking. We validated a method to score clot density in quarter milk samples and describe the prevalence and dynamics of the density scores between consecutive samplings and periods in 4 farms with automatic milking systems. Using in-line filters, we collected quarter milk samples at each milking during 3 periods of 30 h each in each farm. Clot density was scored based on coverage of the filter area as 0 (negative), 1 (trace), 2 (mild), 3 (moderate), 4 (heavy), and 5 (very heavy). The score for a specific quarter and milking is referred to as the quarter milking score (QMS). Three assessors independently scored 902 images of filter samples with a Fleiss kappa value of 0.72. In total, 21,202 quarter milk samples from 5,398 milkings of 621 cows were collected. Of the quarter filter samples, 2.4% had visible clots, distributed as mild (1.4%), moderate (0.6%), heavy (0.3%), and very heavy (<0.1%, n = 8). Cases with a cow period sum of QMS ≥ 4, corresponding to 9.4% of all periods, harbored 86% and 94% of all QMS of 2 to 5 and 3 to 5, respectively. Of these cases, cows sampled in all 3 periods and clots in only 1 period had a quarter period sum score ≥ 1 in 1.8 different quarters in average. Corresponding numbers for the cows with clots or traces in 2 or 3 periods were 2.2 and 2.5 different quarters, respectively. A QMS of 2 to 5 in the preceding milking increased the chance of a QMS >1 in the following milking, with an average chance of 38%. The probability of a QMS > 1 increased with increasing previous QMS, a higher sum of QMS during the milking period, longer milking interval, and higher lactation number, but decreased with increasing days in milk. Our study showed that the method of clot-density scoring is feasible to perform and reproducible for investigating the occurrence and dynamics of clots in milk. Elevated clot-density scores clustered within certain cows and cow periods and appeared in new quarters of the cows over time. The low recurrence of QMS of 1 and 2 within quarters indicated that QMS 3 could be a reasonable threshold for detecting quarters with abnormal milk that require further attention.

Sensitivity and Specificity for the Detection of Clinical Mastitis by Automatic Milking Systems in Bavarian Dairy Herds

“…For example, it has been found that cows with longer milking intervals in DeLaval and GEA systems have a lower chance of being correctly classified as healthy cows. This could be because the likelihood of a cow having abnormal milk increases as the milking interval increases [ 74 ], and the flakes that may be present may affect the measurement of EC [ 18 ], making it difficult to alert correctly; additionally, a milking visit could be protracted due to pain caused by the onset of udder inflammation.…”

Detecting and predicting changes in milk homogeneity using data from automatic milking systems

“…The presence of clots in 2 out of 3 consecutive milkings has been suggested to be included in the "gold-standard" definition of CM (Mein and Rasmussen, 2008;Kamphuis et al, 2013Kamphuis et al, , 2016. The likelihood of a cow having a milking with clots decreases with shorter milking intervals and increases with longer milking intervals (Hallén Sandgren et al, 2021). Thus, the standard of considering clot presence in 2 out of 3 consecutive milkings as indicating abnormal milk might be somewhat misleading for AMS with large variations in milking intervals.…”

“…Although the farmer would not likely detect or take notice of these cases, the traces might give extra information to the algorithm because observations with QMS = 1 seemed to accumulate to a larger extent among cows having clots in the milk (i.e. QMS ≥ 2) than in cows without clots (Hallén Sandgren et al, 2021) and they were therefore included in the CPC models. However, the size of a flake is hard to judge (Rasmussen, 2005), and many observations of traces might have been of single flakes.…”

“…The test of detection systems for changes in milk homogeneity (ISO, 2007) states that clots larger than 2 mm should be considered abnormal in both quarter and composite testing, whereas the density of clots in the filter, which might be a more appropriate way of judging the severity of the change in milk homogeneity, is not mentioned. Possibly, QMS < 3 should also be considered "non-cases," also strengthened by the low repeatability of QMS 1 and 2 within period described in Hallén Sandgren et al (2021). In one of the potential gold standard definitions investigated by Claycomb et al (2009), low-density scores at the cow composite level were excluded, which increased the Se.…”

In-Line Detection of Clinical Mastitis by Identifying Clots in Milk Using Images and a Neural Network Approach

“…A similar sensor has been proposed in the past but was limited in its capabilities of detecting the clots on the filter, as it was developed to score the quality of the milk and needed to be adapted for instances of different detriments occurring in the milk [18]. Another study proposed to measure clot density in quarter milk samples, which could be useful in monitoring milk quality and clinical mastitis [19]. The researchers used in-line filters to collect quarter milk samples and visually scored the clot density, based on the coverage of the filter area.…”