Microbial contamination in drinking water supplies poses ongoing challenges, even in developed countries (Hrudey and Hrudey, 2019). Detecting causal pathogens rapidly remains challenging, and current online methods rely on surrogate indicators, such as turbidity and pH. Emerging technologies, like flow cytometry (FCM), offer high-temporal-resolution monitoring, providing insights into microbial communities. Here, we explore the need for real-time microbial anomaly detection. We also introduce a novel computational model, Microbial Community Change Detection (MCCD), designed to transform microbial community characteristics into an online process control signal.
The Power of Flow Cytometry
The flow cytometry method (FCM) has gained prominence in water quality monitoring, offering detailed insights into microbial dynamics (Egli and Stefan, 2015; Van Nevel et al., 2017b; Safford and Bischel, 2018). Traditional methods, like heterotrophic plate counts, have limitations in speed (2-7 days). FCM's ability to measure suspended particles in real time makes it a valuable tool for assessing microbiological quality during water treatment.
Challenges in Microbial Concentration Monitoring
However, microbial concentration in water processes can fluctuate significantly due to various factors (Besmer and Hammes, 2016; Buysschaert et al., 2018b; Schleich et al., 2019). Periodic fluctuations, biofilm growth, and seasonal changes can lead to challenges in distinguishing normal from abnormal microbial changes.
Beyond Cell Concentration: Cytometric Fingerprints
Flow cytometric measurements provide cell concentration and cytometric fingerprints, summarizing bacterial distribution in the signal space (Koch et al., 2014). While offline methods have been used for fingerprint analysis, there are innovative online approaches for real-time surveillance, such as BactoSense.
Introducing MCCD: A Real-Time Anomaly Detection Model
The MCCD model addresses the need for automatic and real-time surveillance of microbial populations in water processes. It utilises a two-step analysis, transforming flow cytometric measurements into simplified fingerprints using the Probability Binning algorithm (Roederer et al., 2001; Rogers and Holyst, 2009). The fingerprints are then fed into an online model that compares them to a collection of reference measurements, enabling the calculation of an outlier score.
Testing MCCD's Performance
To assess MCCD's performance, we conducted in silico and in vitro tests, simulating acute contaminations in real-world water systems. The model demonstrated robustness against dynamic variations while reliably detecting intentional contaminations.
Conclusion
In conclusion, the MCCD model, with its ability to provide an online outlier score, offers a promising solution for quickly detecting potential microbiological contamination in water systems. This becomes especially crucial when the time between treatment and distribution is minimal. By leveraging FCM's capabilities and advanced computational models, the water industry can enhance its ability to ensure safe and high-quality drinking water.
Read the full Frontiers article here.
Comments