Publikationsdatenbank

Lameness detection in dairy cows by logistic regression model with mixed effects based on accelerometer data from six farms in Germany (2021)

Art

Poster

Autoren

Lavrova, Anastasia (WE 16)

Palmini, Andrea (WE 16)

Choucair, Alexander (WE 18)

Stock, Kathrin F. (WE 11)

Kammer, Martin

Querengässer, Friederike (WE 16)

Doherr, Marcus (WE 16)

Müller, Kerstin-Elisabeth (WE 18)

Belik, Vitaly (WE 16)

Kongress

27. DACH-Epidemiologietagung

Bern, Schweiz, 01. – 03.09.2021

Quelle

27. DACH-Epidemiologietagung „Epidemiologie in der ökologischen Landwirtschaft“ : Gemeinsame Tagung des Forums für Epidemiologie und Tiergesundheit Schweiz der DVG-Fachgruppe „Epidemiologie und Dokumentation“ Institut für Öffentliches Veterinärwesen der Vetmeduni Wien in Verbindung mit der Vetsuisse Fakultät der Universität Bern : 1. bis 3. September 2021 — Veterinary Public Health Institut (Hrsg.)

— S. 60

Sprache

Englisch

Verweise

URL (Volltext): http://dachepi.vphibern.ch/wp-content/uploads/2021/08/Tagungsband_DACH-Epidemiologietagung2021.pdf

Kontakt

Institut für Veterinär-Epidemiologie und Biometrie

Königsweg 67
14163 Berlin
+49 30 838 56034
epi@vetmed.fu-berlin.de

Abstract / Zusammenfassung

Lameness in dairy cows is one of major challenges on the way of improving animal well-being and optimizing economic efficiency. A promising approach for automated animal surveillance for early lameness detection and prevention utilizes cow activity sensors [1]. In the present study we analyzed activity (accelerometer), additional cow-individual as well as farm-related indicators for 3 746 Holstein dairy cows which were scored for lameness for ca. 29 500 times from a longitudinal study during 2015-2016 in six farms in Germany. We developed a statistical model (logistic regression with mixed effects) able to detect lameness in dairy cows with 86% sensitivity and 82% specificity. The following statistically significant independent variables were taken into account: number of steps, mean lying bout duration, mean daily milk yield, days in milk, parity, seasonality. Also their interactions were considered. The parity turned out to be the most important predictor for lameness (OR = 2.1, 95%CI = (1.9,2.4) and p-value= 0.01). Days in milk turned out to be slightly significant (p-value= 0.028) with OR = 0.998, 95%CI = (0.996,1.000). The probability to be lame significantly decreases with the daily milk
yield increase – from ca. 50% for low milk yield (5 kg) to 25% for high milk yield (50 kg). Also, seasonality is statistically significant for the prediction of lameness. The number of steps and the lying bout duration are statistically significant (p-values< 0.001) and negatively influence the probability of lameness. Our further research aims at the implementation of advance machine learning methods for lameness detection, such as deep learning [2] and boosted regression trees [3]. Our results show the potential of automated animal surveillance and promise to significantly improve lameness detection in dairy livestock.
Authors are supported by funds of the federal Ministry of Food and Agriculture (BMEL) based on decision of the Parliament of the Federal Republic of Germany via the Federal Office for Agriculture and Food (BLE) under the innovation support program within Klauenfitnet 1.0 and Klauenfitnet 2.0 consortia.