Near-infrared spectroscopy analysis of blood plasma for predicting nonesterified fatty acid concentrations in dairy cows

Guilherme L Menezes; Tiago Bresolin; Rafael Ferreira; Henry T Holdorf; Sebastian I Arriola Apelo; Heather M White; JoaoR R Dórea

doi:10.3168/jdsc.2023-0458

Near-infrared spectroscopy analysis of blood plasma for predicting nonesterified fatty acid concentrations in dairy cows

JDS Commun. 2023 Nov 17;5(3):195-199. doi: 10.3168/jdsc.2023-0458. eCollection 2024 May.

Authors

Guilherme L Menezes¹, Tiago Bresolin², Rafael Ferreira¹, Henry T Holdorf¹, Sebastian I Arriola Apelo¹, Heather M White¹, JoaoR R Dórea^{1

3}

Affiliations

¹ Department of Animal and Dairy Sciences, University of Wisconsin-Madison, Madison, WI 53706.
² Department of Animal and Dairy Sciences, University of Illinois Urbana-Champaign, Urbana, IL 61801.
³ Department of Biological Systems Engineering, University of Wisconsin-Madison, Madison, WI 53706.

Abstract

During the transition period, dairy cows are often exposed to negative energy balance (NEB), leading to lipid mobilization from adipose tissue into nonesterified fatty acids (NEFA), a common indicator of heightened illness risk. This study aimed to use blood near-infrared (NIR) spectra data to classify NEB into high or low categories, based on early-lactation cow NEFA thresholds. We collected a total of 186 plasma samples from 100 Holstein cows. The samples were categorized into critical thresholds, based on previous literature, of ≥0.60 and ≥0.70 mEq/L for identifying high NEB. Spectral data were preprocessed before the development of the predictive modes, which included the implementation of multiplicative scatter correction, standard normal variate (SNV), and first and second derivatives. The classification was performed using partial least square discriminant analyses (PLS-DA), and predictive performance was assessed using leave-one-out cross-validation. Predictive quality for each class was evaluated through specificity, precision, sensitivity, and F₁ score. The study showed promising results, with the SNV technique achieving higher F₁ scores. The model found 72.7% specificity, 78.9% precision, 80.8% sensitivity, and 79.8% F₁ score to classify animals with NEFA levels of ≥0.60 mEq/L, and 82.1% specificity, 78.7% precision, 80.8% sensitivity, and 79.7% F₁ score to classify animals with NEFA levels ≥0.70 mEq/L. These results indicate that NIR spectroscopy could serve as a tool for detecting cows under severe NEB, also showing potential for broader application across the entire transition period, as the spectral signal carried relevant information regarding cow metabolism. Furthermore, the combination of predictors derived from plasma spectra and other cow-level information can lead to more accurate disease alerts, given their relationship with the NEB.