Adsa Logo White Adsa Title White

Using activity and rumination data to early predict anaplasmosis in dairy calves.

T. Bresolin



Join T. Bresolin on this page for a live text chat!
6:00 PM - 8:00 PM GMT


Using activity and rumination data to early predict anaplasmosis in dairy calves.
V. A. Teixeira1, G. M. Souza2, L. D. Ferreira1, T. R. Tomich3, T. Bresolin*4, M. M. Campos3, A. M. Q. Lana1, S. G. Coelho1, J. A. G. Silveira1, A. U. Carvalho1, E. J. Facury Filho1, L. G. R. Pereira3,4, J. R. R. Dorea4. 1Universidade Federal de Minas Gerais Belo Horizonte, MG, Brazil, 2Universidade Federal de Lavras Lavras, MG, Brazil, 3Brazilian Agricultural Research Corporation � Embrapa Juiz de Fora, MG, Brazil, 4University of Wisconsin Madison, WI.

Bovine anaplasmosis causes large economic losses in dairy cattle production systems worldwide and is commonly detected through rectal temperature, blood smears under a microscope, and packed cell volume (PCV). Such methodologies are labor, costly, and difficult to apply in large scale operations. The objective of this study was to investigate the feasibility of using activity and rumination data retrieved from the SCR Heatime HR collar to identify dairy calves with anaplasmosis. Thirty-four calves with an average age of 119 � 15 d (148 � 20 kg of BW) were challenged with 2 � 107 erythrocytes infected with UFMG1 strain (GenBank no. EU676176) isolated from Anaplasma marginale. After the challenge-exposure, animals were monitored every day by assessing PCV. A PCV critical level of 16 � 3.6% was used as a threshold to classify the animals as sick (0 d) and to initiate an enrofloxacin treatment. Two-time series (TS) were built using the activity and rumination data sets. The first TS (TS1) included the last sequence of 8 d (−8 d to −1 d) preceding the PCV critical level and the second TS (TS2) comprised a sequence of 8 d randomly selected in a window from −50 to −15 d before the PCV critical level. TS2 was randomly selected within the animal to ensure a sequence of days in which PCV was considered normal (32 � 2.4%). Therefore, TS1 and TS2 were assumed to be a sequence of days resulting in critical and non-critical PCV levels, respectively. A Recurrent Neural Network (RNN) approach implemented in Keras R package was used to analyze activity and rumination data sets. To validate the models, leave-one-out cross-validation was performed by removing a pair of time series (TS1 and TS2) per animal at a time. RNN predicted anaplasmosis animal infection from rumination data with accuracy, sensitivity, specificity, positive predictive value, and negative predicted value of 84, 76, 91, 90, and 79%, respectively, whereas for activity these values were 74, 71, 76, 75, and 72%, respectively. In conclusion, activity and rumination data from wearable sensors can potentially be used as early predictors of bovine anaplasmosis in dairy calves.

Keywords: precision dairy farming, rickettsia, tick fever.