Handheld near-infrared spectroscopy for detecting staphylococcal inoculation-associated changes in ex vivo human bone

Bone infections remain diagnostically challenging because current standard methods rely on microbiological and histopathological testing, which are time-consuming and may delay treatment decisions. This study evaluated handheld near-infrared spectroscopy for detecting spectral changes associated with staphylococcal inoculation in ex vivo human bone. An ex vivo bacterial inoculation model was established using human trabecular bone specimens inoculated with Staphylococcus aureus and Staphylococcus epidermidis. Spectra were acquired using a handheld NIR spectroscopy platform and analysed after standard normal variate preprocessing combined with either smoothing or first-derivative transformation. In total, 120 averaged spectra derived from 40 donors were included. Principal component analysis revealed partial clustering by inoculation status, particularly after derivative preprocessing, but substantial overlap remained. Linear discriminant analysis indicated stronger discrimination between inoculated and uninoculated bone than between the two staphylococcal species. However, classification performance depended on preprocessing and validation strategy; therefore, donor-wise dataset partitioning was used to reduce the risk of overly optimistic performance estimates caused by donor-level information leakage. These findings suggest that handheld NIR spectroscopy may provide rapid, non-destructive information on inoculation-associated spectral alterations in controlled ex vivo bone models. Further studies with sham-incubated controls, repeated donor-level cross-validation, independent external validation, additional pathogens, and clinically realistic confounders are required before clinical translation.