Gas-Phase Detection of UTI-Causing Bacteria Using Off-the-Shelf Gas Sensors and Change-Point Detection

Urinary tract infections (UTIs) are bacterial infections affecting the urinary tract, including the kidneys, bladder, and urethra. It’s a common infection in the general population, with women at higher risk. Current diagnostic methods, like urine dipstick tests and pathogen cultures have limitations. Dipstick tests are fast but imprecise. Cultures are accurate but time-consuming, leading to delays in diagnosis.

Gas-phase detection offers a promising alternative, as bacterial growth releases volatile organic compounds (VOCs) like ethanol, acetone, and ammonia. Metal-oxide (MOX) gas sensors provide a rapid, cost-effective way to detect these emissions, enabling real-time monitoring of bacterial proliferation. While they lack molecular specificity, previous experiments have demonstrated their effectiveness in identifying bacterial growth, including Escherichia coli on blood agar.

The proposed detection method improves gas-phase bacterial detection by integrating change-point detection (CPD) with MOX sensor technology, building on previous research. Bacterial VOC emissions were modeled as exponential due to microbial metabolism, gas diffusion, and adsorption kinetics. The isolate-detect (ID) methodology and a slope-threshold algorithm were applied to enhance detection accuracy and reduce false negatives.

The BME688 sensor, operating at 300°C, detects VOCs, volatile sulfur compounds (VSCs), carbon monoxide (CO), and hydrogen by measuring resistance changes caused by oxygen adsorption and gas interactions. Integrated with an ESP32 Pico D4 kit and a custom PCB, it enables real-time monitoring through a MATLAB interface, while simultaneously collecting data.

Bacterial suspensions (10²–10⁶ cfu/mL) were prepared using the Den-600 densitometer to cover values above and below the 10⁴ cfu/mL infection threshold. Petri dish lids with gas sensors were sterilized, air-dried, and preconditioned in the incubator for one hour before data collection.

Serial dilutions were streaked onto blood agar plates, inverted over the gas sensors at 37°C ± 1°C, and tested across three independent trials. Environmental conditions, including temperature, pressure, and humidity, were monitored throughout the experiment. Humidity stabilized within 30 to 40 minutes, while the sensor’s on-chip heater accelerated temperature equilibration.

Experiments showed that the resistance varied with bacterial concentration—higher concentrations led to earlier and stronger resistance shifts, while excessive VOC emissions at high levels caused sensor saturation, marked by a secondary resistance peak followed by an exponential decline.

An offline algorithm analyzed gas-phase measurements, applying Gaussian noise to enhance performance and reduce artifacts. A threshold-based decision rule identified key change points marking bacterial proliferation. A steep downward shift in sensor resistance reliably distinguished infection-level concentrations (≥10⁴ cfu/mL) from lower "subinfection" cases.

For E. coli ATCC25922, diagnostic change points correlated with bacterial concentrations above 10⁴ cfu/mL, confirming the algorithm's predictive potential. Similar results were observed for other bacterial species like E. faecalis, S. agalactiae, and P. aeruginosa, with infection-level concentrations consistently triggering change points within 10 hours.

In conclusion, gas-phase detection with BME688 sensors and CPD-based analysis offers a fast, cost-effective alternative to traditional UTI diagnostics by detecting infection-level concentrations more efficiently. The sensor has potential for clinical diagnostics, point-of-care testing, and antimicrobial resistance monitoring, with future research focused on expanding strain testing and improving real-time bacterial detection.