The preliminary assessment of using the artificial neural networks to diagnose ketosis in Polish Red cattle based on β-hydroxybutyric acid and haematological parameters.

Polish Red (PR) cows belong to an autochthonous conservation breed, kept mainly in the sub-mountain regions of southern Poland. The breed is characterized by good health, but it is susceptible to metabolic disorders such as ketosis. This type of metabolic disease is primarily diagnosed through analyses of β-hydroxybutyric acid (BHB) in milk or blood. Haematological parameters during ketosis might have change and also, be crucial for detecting ketosis as a metabolic disorder. However, limited research study has examined how these parameters change during ketosis onset. The aim of this preliminary study was to use the BHB and haematological parameters to detect cows at risk of subclinical ketosis in PR cattle breed using an artificial neural network (ANN) Multi-Layer Perceptron (MLP) model. The 45% of the animals analysed had elevated plasma BHB concentrations ranging from 0.86 to 6.7 mmol/L. The sensitivity analysis showed satisfied relevance for haematological parameters like lymphocyte count, haemoglobin levels, and blood cells. The effects of the MLP model were verified by the area under the curve (AUC), which was equal to AUC = 0.633 for lymphocytes, AUC = 0.631 for red blood cells volume, and AUC = 0.616 for haemoglobin. Therefore, the novel findings prove the haematological parameters are accurate to build the MLP-ANN model. However, a new bigger data set, and more information criteria are needed in the future work to provide guidance on the choice of the most appropriate MLP-ANN architecture to detect cows with ketosis.