Image created by Dr. Michael J. Miller
Scientists from Kazan Federal University (KFU) and Saint Petersburg Electrotechnical University (LETI) have demonstrated how hyperspectral analysis, an "optical fingerprint" technology, is capable of determining the species composition of bacterial biofilm communities in real-time and detecting the presence of dangerous microorganisms within them. This capability is key to the effective treatment of infections. The study is presented in Analytica Chimica Acta.
The vast majority of infectious diseases in humans and animals are linked, to varying degrees, to the formation of biofilms—complex structures in which communities of microorganisms are immersed in a protective matrix they secrete. This extracellular polymeric substrate creates a reliable shelter for bacteria, making them invulnerable to the immune system and resistant to antimicrobial drugs. It is this resistance that underlies the chronicity of many infections and frequent therapeutic failures.
A particular challenge is that under natural conditions, whether on an implant surface or in a wound bed, biofilms are formed not by a single species, but by several types of microbes. In such mixed consortia, complex interactions arise between bacteria that can radically alter the properties of the entire community. For instance, a previous study showed that a tandem of the fungus Candida and Staphylococcus aureus demonstrates increased resistance, whereas a pairing of S. aureus and Pseudomonas aeruginosa can become more vulnerable to certain antibiotics. Therefore, for effective treatment, it is critically important to quickly and accurately determine exactly who comprises the microbial alliance.
The main goal of the work was to evaluate the potential of hyperspectral analysis for the non-invasive diagnosis of biofilm composition. This technology allows for the acquisition of images of light reflected from an object simultaneously across a multitude of spectral channels, creating its unique "optical passport" without the need for microbiological culture and the isolation of individual microorganism strains.
"The mechanisms underlying the species specificity of the hyperspectral profile are not yet fully clear. We do not yet know exactly what drives this—whether it is primarily the chemical composition of the extracellular polymeric matrix, or if the morphological characteristics of the biofilm play the defining role. Most likely, the observed effect is complex in nature, and the precise mechanisms remain to be elucidated," commented Airat Kayumov, Chair of the KFU Department of Genetics.
Biofilm modeling and biochemical analysis of their matrix were performed at Kazan University with support from the Russian Science Foundation. Hyperspectral measurements and data analysis, including the use of deep machine learning methods, were conducted at LETI. The experiments confirmed that each studied biofilm possesses an individual reflection spectrum. Importantly, the spectrograms of mixed communities are not a simple sum of the spectra of individual species, but rather reflect their complex non-linear interaction, while preserving features that allow for the identification of the consortium participants.
"To identify the species composition of microorganisms in biofilms, we applied a wide spectrum of modern statistical methods—from Bayesian networks, which allow for the visual illustration of correlation links between the spectral structure of reflected light and the biochemical characteristics of the biofilm (allowing a biologist or physician-researcher to interpret the results to some degree), to convolutional networks that integrate spectral and structural information. Through this, accuracy levels of 0.96–0.99 were achieved for certain typical pairs of microorganisms," noted Mikhail Bogachev, one of the study's authors and Chief Researcher at the Department of Radio Engineering Systems LETI.
The new data pave the way for the creation of rapid diagnostic devices. By directing a compact hyperspectral scanner at a wound or implant surface, a doctor could almost instantly (within minutes, compared to several hours for PCR diagnostics and several days for classical bacteriological culture) obtain information on the species composition of the biofilm and select targeted therapy taking into account specific interactions within the microbial community. The path from laboratory prototype to clinical practice requires solving a number of tasks. The main question involves the method's sensitivity under real-world conditions.
"Theoretically, in a controlled environment, these technologies allow for the detection of 10 or more microorganism cells. Therefore, the key task now is testing on real biological surfaces: skin, mucous membranes, as well as on objects contaminated with organic matter. This is what will determine whether the detection threshold corresponds to clinically significant concentrations at the stage of early colonization," added Dr. Kayumov.
Reference
Mikhail I. Bogachev, Pavel S. Baranov, Aleksandr M. Sinitca, Anna V. Mironova, Dmitry R. Sharivzyanov, Alexander A. Basmanov, Elena Y. Trizna, Anna S. Gorshkova, Nikita S. Pyko, Airat R. Kayumov, Non-contact identification of opportunistic pathogens in mixed biofilm contaminations by hyperspectral imaging, Analytica Chimica Acta, Volume 1388, 2026, https://doi.org/10.1016/j.aca.2026.345098.
Abstract
Background
Biofilms are present on almost all surfaces in households, healthcare and medical equipment, foods, industrial and water supply systems, and often contain opportunistic pathogens that represent one of the key public health hazards. The highest risks are imposed by ESKAPEE pathogens, as well as mixed bacterial-fungal communities often exhibiting multiple drug resistance, this way challenging public healthcare.
Results
Here we show how recent developments in hyperspectral imaging technology, complemented by advanced image analysis and machine learning methods, create a unique framework for the potential design of non-contact monitoring systems operating at the macroscale. We could successfully identify five key pathogenic bacteria and a common pathogenic yeast, C. albicans, that frequently occur on surfaces in monospecies and mixed biofilms consisting of combinations of various strains using hyperspectral imaging in the visible, near-infrared, and short-wave infrared spectral bands. Our results indicate that the above pathogenic species could be identified in monocultural biofilms with 95–99.5 % accuracy, while in more frequently occurring mixed biofilms obtained by combining different microorganisms, the accuracy ranges from 90 to 92 % for gram-negative E. coli, K. pneumoniae, and P. aeruginosa to 96–99 % for fungi and gram-positive E. faecalis and S. aureus, respectively, under in vitro conditions.
Significance
Based on our results, we believe that hyperspectral imaging represents a promising and highly efficient technology for the rapid, regular, non-contact monitoring of various equipment and surfaces to detect bacterial and fungal pathogens in situ.
Provided by Kazan Federal University