Variables of gait inconsistency outperform basic gait variables in detecting mildly lame cows

Abstract

a b s t r a c tTo support herdsmen in finding the lame cows on their herds, several automated systems that measure lameness related cow features such as gait patterns, are being developed. Most of these systems are able to distinguish between non-lame and severely lame cows. Detecting mildly lame cows in an early stage of lameness however seems challenging. Inspired by the approach used in human gait research, new variables that measure the inconsistency in stride-to-stride variables were tested using cow gait and were able to show differences between a group of non-lame and a group of mildly lame cows. In order to investigate the added value of these inconsistency variables in detecting mildly lame cows, two new lameness detection models were build: one using solely basic gait variables and a second model using both basic and the new gait inconsistency variables. The second model using the gait inconsistency variables outperformed the model based on only basic gait variables by far in detecting the mildly lame cows with a sensitivity of 88% and a specificity of 87%. These results support the suggestion of incorporating such gait inconsistency variables into lameness detection models. Further validation of these gait inconsistency variables should be investigated using longitudinal studies where cows developing lameness and recovering from it are monitored daily.

Invited review: Cattle lameness detection with accelerometers

“…Van Nuffel et al (2015c) reported that stride length (meters) and duration (seconds) were indicative of lameness using the Gaitwise pressure mat system, a research tool described in Maertens et al (2011). Using the Gaitwise system, stance time (weight-bearing) for the nonlame leg was also found to be longer in lame cows (Van Nuffel et al, 2013, 2015a. Lame cows are relatively cautious or careful about affected foot placement as this action is thought to be painful (Van De Gucht et al, 2017b).…”

“…These variables were generally averages. Measures of inconsistency have also been proposed by some researchers citing its use in human gait research (Van Nuffel et al, 2015a). Inconsistency in this context is the extent a step varies from one to the next.…”

“…Inconsistency in this context is the extent a step varies from one to the next. Inconsistency has been operationalized as the coefficient of variation (standard deviation divided by the average value; Van Nuffel et al, 2015a). In that study, most of the studied inconsistency measures had incremental predictive value over the basic variables.…”

A Framework for Real-time Cattle Monitoring using Multimedia Networks

“…Mobility scoring is one example of a subjective technique used by herdsmen to identify lameness and locomotors problems in dairy cows [7]. Although cheap to disseminate, mobility scoring is time-consuming and must be done regularly [9][10]. Another criticism of the technique is that it may fail to identify the early (sub-clinical) stages of lameness [11][12].…”

Farm-specific economic value of automatic lameness detection systems in dairy cattle: From concepts to operational simulations

“…On top of that, lack of application of visual locomotion scoring in practice often results in bad diagnoses, untreated lameness, and associated losses. To solve this problem, research started on the development of automatic lameness detection systems using a wide range of sensor techniques such as cameras (Nikkhah et al, 2005;Viazzi et al, 2013;Van Hertem et al, 2014), pressure plates (Neveux et al, 2006;Chapinal et al, 2009), pressure mats (Maertens et al, 2011;Van Nuffel et al, 2015), and accelerometers (Chapinal et al, 2009;Pastell et al, 2009). Research, however, has principally focused on sensor development and data interpretation (Rutten et al, 2013), whereas studies on the economic added value of such systems for farmers are scarce.…”