Portable near-infrared spectroscopy: A rapid and accurate blood test for diagnosis of Haemonchus contortus infection and for targeted selective treatment of sheep

[Display omitted] •Portable near-infrared spectroscopy (NIRS) was used for diagnosis of Haemonchus contortus infection in lambs.•NIRS showed high sensitivity and specificity in the diagnosis of H. contortus.•The partial least squares method showed high precision and accuracy to estimate the sheep packed cell volume values.•NIRS showed better results in comparison with the FAffa MAlan CHArt (FAMACHA) for detection of lambs infected.•NIRS could be a tool for the future performance of targeted selective treatment. Haemonchus contortus is the most prevalent and important gastrointestinal nematode (GIN) in small ruminants. Since it reduces the packed cell volume (PCV), causing anemia, early diagnosis can be used for targeted selective treatment (TST) of sheep, reducing antiparasitic drug use and anthelmintic resistance. This study aimed to predict PCV values through near-infrared spectroscopy (NIRS) and to develop a classification and diagnosis model of H. contortus infection using PCV values, eggs per gram of feces (EPG) counts and mean daily weight gain (DWG). A total of 1728 spectra were collected from blood samples of 216 lambs with a portable NIR spectroscope. In parallel, other parameters indicative of infection were measured: PCV by hematocrit, FAffa MAlan CHArt (FAMACHA) scores, EPG and DWG. To evaluate the relationship between NIRS spectra and the evaluated parameters, principal component analysis (PCA) was used for an exploratory analysis, regression by the partial least squares method (PLS) for the prediction of PCV values via NIRS, and PCA linear discriminant analysis (PCA-LDA) as a classification model for diagnosis. The absorption peaks in the NIRS region associated with the excitation of overtones of nitrogen-hydrogen (N-H) functional groups of proteins had a strong impact on the principal components (PCs), indicating that blood proteins, especially hemoglobin, can be estimated by the NIRS technique. The model for predicting PCV by PLS presented a standard error of prediction of 2.53, root-mean-square error of 2.48, and coefficient of determination of 0.84, indicating good correlation between the PCV values predicted by the model and the PCV obtained by hematocrit. The PCA-LDA model presented 93.33% sensitivity and 82.18% accuracy, both higher than those of the FAMACHA method, as was expected for resilient Morada Nova lambs. The multivariate models associated with the NIRS technique reported here can be used in the future as a quick and versatile