| Autor: |
Galindo C; Institute of Applied Physics, The Hebrew University of Jerusalem, Jerusalem 9190401, Israel., Levy G; Institute of Applied Physics, The Hebrew University of Jerusalem, Jerusalem 9190401, Israel., Feldman Y; Institute of Applied Physics, The Hebrew University of Jerusalem, Jerusalem 9190401, Israel., Roth Z; Animal Science Department, The Robert H. Smith Faculty of Agriculture, Food, and Environment, The Hebrew University of Jerusalem, Rehovot 7610001, Israel., Shalev J; Animal Science Department, The Robert H. Smith Faculty of Agriculture, Food, and Environment, The Hebrew University of Jerusalem, Rehovot 7610001, Israel., Raz C; Animal Science Department, The Robert H. Smith Faculty of Agriculture, Food, and Environment, The Hebrew University of Jerusalem, Rehovot 7610001, Israel., Mor E; The Racah Institute of Physics, The Hebrew University of Jerusalem, Jerusalem 9190401, Israel., Argov-Argaman N; Animal Science Department, The Robert H. Smith Faculty of Agriculture, Food, and Environment, The Hebrew University of Jerusalem, Rehovot 7610001, Israel. |
| Abstrakt: |
The most reliable methods for pregnancy diagnosis in dairy herds include rectal palpation, ultrasound examination, and evaluation of plasma progesterone concentrations. However, these methods are expensive, labor-intensive, and invasive. Thus, there is a need to develop a practical, non-invasive, cost-effective method that can be implemented on the farm to detect pregnancy. This study suggests employing microwave dielectric spectroscopy (MDS, 0.5-40 GHz) as a method to evaluate reproduction events in dairy cows. The approach involves the integration of MDS data with information on milk solids to detect pregnancy and identify early embryonic loss in dairy cows. To test the ability to predict pregnancy according to these measurements, milk samples were collected from (i) pregnant and non-pregnant randomly selected cows, (ii) weekly from selected cows ( n = 12) before insemination until a positive pregnancy test, and (iii) daily from selected cows ( n = 10) prior to insemination until a positive pregnancy test. The results indicated that the dielectric strength of Δ ε and the relaxation time, τ , exhibited reduced variability in the case of a positive pregnancy diagnosis. Using principal component analysis (PCA), a clear distinction between pregnancy and nonpregnancy status was observed, with improved differentiation upon a higher sampling frequency. Additionally, a neural network machine learning technique was employed to develop a prediction algorithm with an accuracy of 73%. These findings demonstrate that MDS can be used to detect changes in milk upon pregnancy. The developed machine learning provides a broad classification that could be further enhanced with additional data. |
