Respiratory illnesses represent a major challenge in modern pig production, leading to notable economic losses and heightened antimicrobial usage. Reliable differentiation of lung lesions is essential for accurate diagnosis and effective disease control. The incorporation of fast, non-invasive technologies could significantly improve the management of such issues. This research explored the feasibility of employing near-infrared (NIR) spectroscopy to classify porcine lung tissue. Spectral data (908–1676 nm) from 101 lungs collected from weaned pigs were obtained using a portable device and analyzed through multivariate statistical techniques. Two distinct discriminant models were designed to distinguish between normal (N), congested (C), and pathological (P) lungs, as well as among catarrhal bronchopneumonia (CBP), fibrinous pleuropneumonia (FPP), and interstitial pneumonia (IP) forms. The model optimized for identifying specific pathological types yielded the highest classification performance. The main limitations emerged with C lungs, which showed a 30% misclassification rate with N and P samples, and with FPP lesions, which were incorrectly recognized as CBP in 30% of cases. Conversely, CBP and IP samples were consistently detected with accuracy, sensitivity, and precision exceeding 90%. In summary, these findings provide proof of concept for using NIR spectroscopy to distinguish and classify porcine lungs with various lesions, supporting future rapid and efficient diagnostic applications.
