Search

Subclinical ketosis is an important and common disease in early lactating dairy cows, associated with economic losses because of decreased milk production, decreased reproductive performance and increased risk of additional periparturient diseases. Subclinical ketosis is defined by increased blood concentrations of ketone bodies in the absence of clinical signs of ketosis. The gold standard for subclinical ketosis is the measurement of BHBA (β- hydroxybutyrate) in serum or plasma. Available on-site blood tests detect the disease with high accuracy, but are expensive and labor-intensive on a herd level. Introducing easier procedures for detecting the disease in practice, or at least reducing the number of animals which have to be blood-tested, by identifying animals at risk, are desired. Our approach is based on the SMARTBOW system, an ear-tag equipped with a 3D accelerometer, which records lying, standing, rumination and activity behavior. Blood samples for the measurement of BHBA were used as gold standard. We propose a new measure called MITOD (Minimal Total Distance), which aims to flexibly compare time series and compensate for possible alignment errors. This distance measure was incorporated in a machine learning approach and it's classification quality was compared with state of the art methods on our data to infer the health status of the considered animals. Our approach shows promising results for detecting subclinical ketosis and for classifying time series in general. This work has been supported by the COMET-K2 Center of the Linz Center of Mechatronics (LCM) funded by the Austrian federal government and the federal state of Upper Austria. Animal health data were collected in the agriProKnow project funded by the Austrian Research Promotion Agency (FFG, project# 848610). [ABSTRACT FROM AUTHOR]