Recently, a team of researchers from Huazhong Agricultural University developed diagnostic models using mid-infrared spectroscopy (MIR) to improve the detection of mastitis in dairy cattle. Their study, which was recently published in Animals, demonstrates how MIR technology can offer a faster, less costly alternative to conventional testing methods, while maintaining strong diagnostic accuracy (1).
Mastitis is an inflammation of the udder typically caused when microbes enter the teat (1,2). This infection is often described in two ways: clinical or subclinical. Clinical mastitis is the inflammatory response to infection that results in fibrin clots and unusual milk coloration (2). Subclinical mastitis is characterized by the absence of local inflammation or systemic involvement, and this generally results in a sharp decrease in milk production (2).
Mastitis is one of the most common and economically damaging diseases in dairy farming worldwide. Its presence leads to reduced milk yield, poorer milk quality, and higher veterinary costs, creating a substantial burden for dairy producers (1). Currently, the main indicators of udder health are somatic cell count (SCC) and differential somatic cell count (DSCC) (1). Although reliable, these indicators are measured through flow cytometry, a method that is expensive, time-consuming, and particularly difficult for widespread implementation when analyzing DSCC.
As part of their experimental procedure, the research team tested whether MIR spectroscopy could be adapted into accurate diagnostic models for udder health, with the goal of replacing or supplementing SCC and DSCC testing on dairy farms. MIR spectroscopy has been used in the assessment of milk composition because it is nondestructive and cheap (1). This study tried to see if MIR spectroscopy could be taken a step further in evaluating udder health.
The study analyzed 2,288 milk samples collected from dairy farms, using a CombiFoss 7 DC instrument manufactured by FOSS (1). The researchers tested three MIR spectral preprocessing methods, six modeling algorithms, and three different sets of MIR spectral data to design several diagnostic models (1).
The results of the study showed that two models were very good for diagnosing udder health. The first model, which was named “DIFF-RF-1060 wavenumbers,” combined a spectral preprocessing method of difference (DIFF), a Random Forest (RF) algorithm, and 1,060 wavenumbers (1). This model successfully distinguished healthy cattle from those with mastitis, achieving an area under the curve (AUC) score of 1.00 in the training set and 0.80 in the test set (1).
The second model, which was called “DIFF-SVM-274 wavenumbers,” paired the DIFF preprocessing method with a Support Vector Machine (SVM) algorithm and 274 wavenumbers (1). This model differentiated cases of mastitis from chronic or persistent mastitis, achieving AUC scores of 0.87 in the training set and 0.85 in the test set (1).
The results demonstrate the significant potential of MIR-based diagnostic models to support dairy farmers with faster and more affordable mastitis detection.
Although the two models demonstrated potential in diagnosing udder health, the researchers acknowledged more work is needed before their method is widely disseminated. They explained that a model designed to distinguish between healthy cattle and those with suspicious mastitis showed poor test performance, indicating that additional research is needed (1). The authors noted that incorporating more representative and diverse samples would likely improve the precision and versatility of the diagnostic models (1). They also suggested future work could integrate factors such as lactation stage data to refine prediction accuracy.
The development of these MIR-based diagnostic tools could mark a turning point in dairy health management. By expanding the application of SCC and DSCC proxies while lowering costs, the models have the potential to streamline mastitis monitoring, reduce economic losses, and improve animal welfare across dairy farms (1).
