Application of machine-learning methods to milk mid-infrared spectra for discrimination of cow milk from pasture or total mixed ration diets

Using milk mid-infrared spectroscopy to estimate cow-level nitrogen efficiency metrics

Minimizing pollution from the dairy sector is paramount; one potential cause of such pollution is excess nitrogen. Nitrogen pollution contributes to a deterioration in water quality as well as an increase in both eutrophication and greenhouse gases. It is therefore essential to minimize the loss of nitrogen from the sector, including excretion from the cow. Breeding programs are one potential strategy to improve the efficiency with which nitrogen is used by dairy cows, but they rely on routine access to individual cow information on how efficiently each cow uses the nitrogen it ingests. A total of 3,497 test-day records for individual-cow nitrogen efficiency metrics along with milk yield and the associated milk spectra were used to investigate the ability of milk infrared spectral data to predict these nitrogen traits; both traditional partial least squares regression and neural networks were used in the prediction process. The data originated from 4 farms across 11 yr. The nitrogen traits investigated were nitrogen intake, nitrogen use efficiency, and nitrogen balance. Both nitrogen use efficiency and nitrogen balance were calculated considering nitrogen intake, nitrogen in milk, nitrogen in the conceptus, nitrogen used for the growth, nitrogen stored in body reserves, and nitrogen mobilized from body reserves. Irrespective of the nitrogen-related trait being investigated, the best predictions from 4-fold cross validation were achieved using neural networks that considered both the morning and evening milk spectra along with milk yield, parity, and DIM in the prediction process. The coefficient of determination in the cross validation was 0.61, 0.74, and 0.58 for nitrogen intake, nitrogen use efficiency, and nitrogen balance, respectively. In a separate series of validation approaches, the calibration and validation was stratified by herd (n = 4) and separately by year. For these scenarios, partial least squares regression generated more accurate predictions compared with neural networks; the coefficient of determination was always lower than 0.29 and 0.60 when validation was stratified by herd and year, respectively. Therefore, if the variability of the data being predicted in the validation datasets is similar to that in the data used to develop the predictions, then nitrogen-related traits can be predicted with reasonable accuracy. In contrast, where the variability of the data that exists in the validation dataset is poorly represented in the calibration dataset, then poor predictions will ensue.

Comparing the potential of benchtop and handheld mid-infrared spectrometers for predicting soil phosphorus (P) sorption capacity and evaluating the influence of sample preparation

Traditional soil phosphorus (P) sorption capacity is examined from a Langmuir isotherm batch technique, which is time-consuming, labour intensive and generates chemical waste. In this work, we provide an efficient and convenient technique with MIR spectroscopy to predict the Langmuir parameter of soil P sorption maximum capacity (Smax, mg·kg-1). Four spectral libraries from benchtop (Bruker) and handheld (Agilent) MIR spectrometers were built with samples in two particle size ranges, <0.100 mm (ball-milled) and <2 mm. respectively. Using an archive of samples with a database of sorption parameters, soils were classified into 'low' and 'high' sorption capacities. Chemometric regression models of partial least squares (PLS), Cubist, support vector machine (SVM) regression and random forest (RF) were evaluated for Smax prediction. Bruker spectral libraries with both soil particle sizes yielded 'excellent models', with SVM predicting Smax values with high accuracy (RPIQV = 4.50 and 4.25 for the spectral libraries of the ball-milled and <2 mm samples, respectively). In comparison, the Agilent handheld spectral libraries contained more noise and less resolution. For Agilent MIR spectroscopy, more homogeneous samples after ball milling resulted in a higher accurate Smax prediction. For Agilent libraries of ball-milled samples, an 'approximate quantitative model' (RPIQV = 2.74) was obtained from the raw spectra using the Cubist algorithm. However, for Agilent spectroscopy of <2 mm samples, the best performing Cubist algorithm can only achieve a 'fair model' (RPIQV=2.23) with the potential to discriminate between 'low' and 'high' Smax values. The results suggest that the benchtop spectrometer can predict the Langmuir Smax value with high accuracy without the need to ball mill samples. However, the handheld spectrometer can only make approximate quantitative predictions of Smax for ball-milled samples. For <2 mm samples, Agilent can only be used to classify 'low' and 'high' sorption capacity soils.

Usefulness of mid-infrared spectroscopy as a tool to estimate body condition score change from milk samples in intensively fed dairy cows

Directly measuring individual cow energy balance is not trivial. Other traits such as body condition score (BCS) and BCS change (ΔBCS) can, however, be used as an indicator of cow energy status. Body condition score is a metric used worldwide to estimate cow body reserves, but the estimation of ΔBCS was, until now, conditional on the availability of multiple BCS assessments. The aim of the present study was to estimate ΔBCS from milk mid-infrared (MIR) spectra and days in milk (DIM) in intensively fed dairy cows using statistical prediction methods. Daily BCS was interpolated from cubic splines fitted through the BCS records and daily ΔBCS was calculated from these splines. The ΔBCS records were merged with milk MIR spectra recorded on the same week. The dataset comprised 37,077 ΔBCS phenotypes across 9,403 lactations from 6,988 cows in 151 herds based in Quebec, Canada. Partial least squares regression (PLSR) and a neural network (NN) were then used to estimate ΔBCS from (1) MIR spectra only, (2) DIM only, or (3) MIR spectra and DIM together. The ΔBCS data in both the first 120 and 305 DIM of lactation were used to develop the estimates. Daily ΔBCS had a standard deviation of 4.40 × 10-3 BCS units in the 120-d dataset and of 3.63 × 10-3 BCS units in the 305-d dataset. A 4-fold cross-validation was used to calibrate and test the prediction equations. External validation was also conducted using more recent years of data. Irrespective of whether based on the first 120 or 305 DIM, or when MIR spectra only, DIM only or MIR spectra and DIM were jointly used as prediction variables, NN produced the lowest root mean square error (RMSE) of cross-validation (1.81 × 10-3 BCS units and 1.51 × 10-3 BCS units, respectively, using the 120-d and 305-d dataset). Relative to predictions for the entire 305 DIM, the RMSE of cross-validation was 15.4% and 1.5% lower in the first 120 DIM when using PLSR and NN, respectively. Predictions from DIM only were more accurate than those using just MIR spectra data but, irrespective of the dataset and of the prediction model used, combining DIM information with MIR spectral data as prediction variables reduced the RMSE compared with the inclusion of DIM alone, albeit the benefit was small (the RMSE from cross-validation reduced by up to 5.5% when DIM and spectral data were jointly used as model features instead of DIM only). However, when predicting extreme ΔBCS records, the MIR spectral data were more informative than DIM. Model performance when predicting ΔBCS records in future years was similar to that from cross-validation demonstrating the ability of MIR spectra of milk and DIM combined to estimate ΔBCS, particularly in early lactation. This can be used to routinely generate estimates of ΔBCS to aid in day-to-day individual cow management.

Estimation of body condition score change in dairy cows in a seasonal calving pasture-based system using routinely available milk mid-infrared spectra and machine learning techniques

Body condition score (BCS) is a subjective estimate of body reserves in cows. Body condition score and its change in early lactation have been associated with cow fertility and health. The aim of the present study was to estimate change in BCS (ΔBCS) using mid-infrared spectra of the milk, with a particular focus on estimating ΔBCS in cows losing BCS at the fastest rate (i.e., the cows most of interest to the producer). A total of 73,193 BCS records (scale 1 to 5) from 6,572 cows were recorded. Daily BCS was interpolated from cubic splines fitted through the BCS records, and subsequently used to calculate daily ΔBCS. Body condition score change records were merged with milk mid-infrared spectra recorded on the same week. Both morning (a.m.) and evening (p.m.) spectra were available. Two different statistical methods were used to estimate ΔBCS: partial least squares regression and a neural network (NN). Several combinations of variables were included as model features, such as days in milk (DIM) only, a.m. spectra only and DIM, p.m. spectra only and DIM, and a.m. and p.m. spectra as well as DIM. The data used to estimate ΔBCS were either based on the first 120 DIM or all 305 DIM. Daily ΔBCS had a standard deviation of 1.65 × 10^-3 BCS units in the 305 DIM data set and of 1.98 × 10^-3 BCS units in the 120 DIM data set. Each data set was divided into 4 sub-data sets, 3 of which were used for training the prediction model and the fourth to test it. This process was repeated until all the sub-data sets were considered as the test data set once. Using all 305 DIM, the lowest root mean square error of validation (RMSEV; 0.96 × 10^-3 BCS units) and the strongest correlation between actual and estimated ΔBCS (0.82) was achieved with NN using a.m. and p.m. spectra and DIM. Using the 120 DIM data, the lowest RMSEV (0.98 × 10^-3 BCS units) and the strongest correlation between actual and estimated ΔBCS (0.87) was achieved with NN using DIM and either a.m. spectra only or a.m. and p.m. spectra together. The RMSEV for records in the lowest 2.5% ΔBCS percentile per DIM in early lactation was reduced up to a maximum of 13% when spectra and DIM were both considered in the model compared with a model that considered just DIM. The performance of the NN using DIM and a.m. spectra only with the 120 DIM data was robust across different strata of farm, parity, year of sampling, and breed. Results from the present study demonstrate the ability of mid-infrared spectra of milk coupled with machine learning techniques to estimate ΔBCS; specifically, the inclusion of spectral data reduced the RMSEV over and above using DIM alone, particularly for cows losing BCS at the fastest rate. This approach can be used to routinely generate estimates of ΔBCS that can subsequently be used for farm decisions.

Importance of Mid-Infrared Spectra Regions for the Prediction of Mastitis and Ketosis in Dairy Cows

Mid-infrared (MIR) spectroscopy is routinely applied to determine major milk components, such as fat and protein. Moreover, it is used to predict fine milk composition and various traits pertinent to animal health. MIR spectra indicate an absorbance value of infrared light at 1060 specific wavenumbers from 926 to 5010 cm−1. According to research, certain parts of the spectrum do not contain sufficient information on traits of dairy cows. Hence, the objective of the present study was to identify specific regions of the MIR spectra of particular importance for the prediction of mastitis and ketosis, performing variable selection analysis. Partial least squares discriminant analysis (PLS-DA) along with three other statistical methods, support vector machine (SVM), least absolute shrinkage and selection operator (LASSO), and random forest (RF), were compared. Data originated from the Austrian milk recording and associated health monitoring system (GMON). Test-day data and corresponding MIR spectra were linked to respective clinical mastitis and ketosis diagnoses. Certain wavenumbers were identified as particularly relevant for the prediction models of clinical mastitis (23) and ketosis (61). Wavenumbers varied across four distinct statistical methods as well as concerning different traits. The results indicate that variable selection analysis could potentially be beneficial in the process of modeling.

Prediction of Indirect Indicators of a Grass-Based Diet by Milk Fourier Transform Mid-Infrared Spectroscopy to Assess the Feeding Typologies of Dairy Farms

This research aims to develop a predictive model to discriminate milk produced from a cattle diet either based on grass or not using milk mid-infrared spectrometry and the month of testing (an indirect indicator of the feeding ration). The dataset contained 3,377,715 spectra collected between 2011 and 2021 from 2449 farms and 3 grazing traits defined following the month of testing. Records from 30% of the randomly selected farms were kept in the calibration set, and the remaining records were used to validate the models. Around 90% of the records were correctly discriminated. This accuracy is very good, as some records could be erroneously assigned. The probability of belonging to the GRASS modality allowed confirmation of the model's ability to detect the transition period even if the model was not trained on this data. Indeed, the probability increased from the spring to the summer and then decreased. The discrimination was mainly explained by the changes in the milk fat, mineral, and protein compositions. A hierarchical clustering from the averaged probability per farm and year highlighted 12 groups illustrating different management practices. The probability of belonging to the GRASS class could be used in a tool counting the number of grazing days.