Gas sensors combined with machine learning can rapidly detect pathogens and antimicrobial resistance by analysing microbial VOCs. This could be a faster, cost-effective alternative to conventional diagnostics.
Image Credit: /Shutterstock.com
A recent report in Cell Biomaterials demonstrates how technology could help clinicians diagnose infections more quickly and accurately. Fast and precise diagnosis is critical in improving patient care and tackling the spread of antimicrobial resistance (AMR).
Analysing Microbial Metabolic Signatures
The study investigates the unique volatile organic compounds (VOCs) emitted by microbes and infected tissues. These compounds, produced as metabolic byproducts, can serve as chemical signatures specific to microbial species and their physiological states, including antibiotic resistance.
VOCs are usually studied using high-precision instruments such as gas chromatography-mass spectrometry (GC-MS) and proton transfer reaction–mass spectrometry (PTR-MS). While effective, these techniques are costly, technically demanding, and unsuitable for point-of-care or bedside use.
Gas sensors, made of materials such as nanostructured metal oxides, conductive polymers, and hybrid composites, offer a more practical alternative. These sensors are compact, affordable, and can detect VOCs or characteristic patterns in real time.
The Role of Machine Learning
Interpreting the complex and often overlapping VOC signals detected by these sensors is a challenge within research. To address this, the study's authors assessed the viability of integrating machine learning models to analyse the multidimensional sensor data and improve diagnostic accuracy.
Their report showcases how algorithms can be successfully applied to classify sensor response patterns. They reviewed support vector machines (SVM), random forests, long short-term memory (LSTM) neural networks, and gradient boosting.
These models can distinguish between bacterial species and between drug-resistant and susceptible strains, even when data are noisy or influenced by environmental variables.
Testing on Cultures, Tissues, and Biofluids
The team compiled findings from recent studies that tested these sensor-ML systems on bacterial cultures, infected tissue samples, and clinical biofluids such as urine and blood.
In these experiments, sensors detected changes in electrical resistance, conductance, or other measurable properties when exposed to VOCs. Well-trained machine learning models could then classify these response patterns, achieving high sensitivity and specificity in identifying bacteria such as Escherichia coli and Staphylococcus aureus.
These techniques can also be extended to identifying antimicrobial resistance profiles. Specific VOC patterns associated with resistant strains, such as those producing extended-spectrum beta-lactamases, were reliably distinguished from susceptible strains.
Towards Portable, Point-Of-Care Diagnostics
The article highlights the ongoing efforts to refine sensor performance, including functionalising sensor surfaces and miniaturising devices for point-of-care applications. It also looks at mitigating environmental factors such as humidity and temperature that can affect results.
The authors note the importance of training machine learning models on datasets that capture the full variability of real-world clinical conditions to ensure high and generalisable performance. Standardising sampling protocols and sensor calibration procedures will also be necessary for their success.
Although further development is needed before these systems can be used in clinical practice, the study shows that portable, user-friendly devices could eventually deliver rapid, noninvasive results in point-of-care settings, complementing existing laboratory methods and supporting better antimicrobial stewardship.
Next Steps and Challenges
The review demonstrates the clear potential of combining AI and sensors for the medical detection of infections. But several hurdles remain. Cross-sensitivity to environmental VOCs and inconsistent sampling can still compromise accuracy.
Clinical validation in diverse settings, as well as establishing standard operating procedures, will be key to translating the technology from laboratory prototypes to widely used diagnostic tools. Integrating advanced sensor arrays with sophisticated machine learning algorithms is critical to achieving the level of accuracy, robustness, and usability required for healthcare.
Download your PDF copy now!
Journal reference
Bilgin M. B., et al. (2025). Microbial and antimicrobial resistance diagnostics by gas sensors and machine learning. Cell Biomaterials. DOI: 10.1016/j.celbio.2025.100125, https://www.cell.com/cell-biomaterials/fulltext/S3050-5623(25)00116-3