Near Infrared Spectroscopy for Rapid Estimation of Somatic Cell Counts in Human Breast Milk

Abstract

Elevated somatic cell counts (SCC) in human milk are associated with mastitis, an inflammation of the breast. However, the presence of fat globules can make the direct measurement of cells in milk challenging. We showed that near infrared (NIR) spectroscopy, a technique that has previously been used in the dairy industry for direct measurement of SCC in bovine milk, can be used for estimating SCC in human milk. Binary classification models were developed using multilinear regression with genetic algorithm searching for selection of wavelets. After correcting NIR frequency spectra for scatter contributions by fat globules and applying a Haar wavelet transform to the data, we found that multivariate classification allowed for separation of samples with low SCC (?150 K cells mL−1) from those with high SCC (?600 K cells mL−1). Sensitivity and specificity for cross-validated NIR estimates were 85% and 84%, respectively. The NIR method had very low rates of misclassification, with a model that used only two wavelets for classification. Additionally, this technique required no sample preparation and has potential as a rapid screening method for identifying elevated SCC in milk of nursing mothers.