Identification and Classification of Multi-Species Biofilms on Polymeric Surfaces Using Hyperspectral Imaging

Abstract

Biofilm-associated contamination poses significant challenges to the food industry, particularly in ensuring effective sanitization and reliable detection. This study explores the use of hyperspectral imaging (HSI) in the shortwave infrared (SWIR) range for non-destructive detection and classification of biofilms on thermoplastic polyurethane (TPU) surfaces. Multi-species biofilms composed of Comamonas sp., Raoultella sp., and Escherichia coli were formed at 10°C and 25°C and biofilm protein and polysaccharide contents were determined. Principal Component Analysis (PCA) and Partial Least Squares Discriminant Analysis (PLS-DA) were used to differentiate among four classes of TPU coupons, including blank (BLANK), control (CTRL), intermediate-strength biofilms formed at 10°C (S10), and strong biofilms formed at 25°C (S25). PCA successfully clustered samples based on spectral profiles of the classes, identifying significant wavelength regions at 1451 and 1926 nm, which correlated with the water, protein, and polysaccharide content of multi-species biofilms. PLS-DA provided a classification accuracy ranging from 68% to 100%, with the highest classification accuracy (100%) observed for BLANK and biofilm-contaminated (S25) TPU coupons and the lowest accuracy (68%) for CTR. Additionally, Partial Least Squares Regression (PLSR) was employed to predict the protein content of biofilms, achieving reliable predictions both in calibration ($R_{\mathrm{cal}}^2$ of 0.81) and external validation ($R_{\text{pred}}^2$ of 0.72). These findings demonstrate the potential of HSI to detect and classify biofilm-infected TPU coupons utilizing wavebands associated with proteins, polysaccharides and water. Hence, HSI can be used as a rapid and non-destructive alternative to traditional methods for biofilm detection, including chemical-based methods such as BioDetect (SANI MARC) and fluorescence-based imaging methods like BACTISCAN.