Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2023 Oct 4;7(10):e2023GH000866.
doi: 10.1029/2023GH000866. eCollection 2023 Oct.

Machine Learning for Detecting Virus Infection Hotspots Via Wastewater-Based Epidemiology: The Case of SARS-CoV-2 RNA

Calvin Zehnder  1 , Frederic Béen  2 , Zoran Vojinovic  1   3   4   5 , Dragan Savic  2   3   4 , Arlex Sanchez Torres  1 , Ole Mark  6 , Ljiljana Zlatanovic  7   8 , Yared Abayneh Abebe  1   9
Affiliations

Machine Learning for Detecting Virus Infection Hotspots Via Wastewater-Based Epidemiology: The Case of SARS-CoV-2 RNA

Calvin Zehnder et al. Geohealth. .

Abstract

Wastewater-based epidemiology (WBE) has been proven to be a useful tool in monitoring public health-related issues such as drug use, and disease. By sampling wastewater and applying WBE methods, wastewater-detectable pathogens such as viruses can be cheaply and effectively monitored, tracking people who might be missed or under-represented in traditional disease surveillance. There is a gap in current knowledge in combining hydraulic modeling with WBE. Recent literature has also identified a gap in combining machine learning with WBE for the detection of viral outbreaks. In this study, we loosely coupled a physically-based hydraulic model of pathogen introduction and transport with a machine learning model to track and trace the source of a pathogen within a sewer network and to evaluate its usefulness under various conditions. The methodology developed was applied to a hypothetical sewer network for the rapid detection of disease hotspots of the disease caused by the SARS-CoV-2 virus. Results showed that the machine learning model's ability to recognize hotspots is promising, but requires a high time-resolution of monitoring data and is highly sensitive to the sewer system's physical layout and properties such as flow velocity, the pathogen sampling procedure, and the model's boundary conditions. The methodology proposed and developed in this paper opens new possibilities for WBE, suggesting a rapid back-tracing of human-excreted biomarkers based on only sampling at the outlet or other key points, but would require high-frequency, contaminant-specific sensor systems that are not available currently.

Keywords: COVID‐19; SARS‐CoV‐2; machine learning; sewer network modeling; support vector machine; wastewater‐based epidemiology.

PubMed Disclaimer

Conflict of interest statement

The authors declare no conflicts of interest relevant to this study.

Figures

Figure 1
Figure 1
Workflow of methodology showing the main processes starting from input data and boundary conditions generation to hydraulic model simulation, signal classification and virus origins identification.
Figure 2
Figure 2
Semi‐hypothetical case‐study pipe network showing service areas, pipes with flow directions, manholes, sampling locations and an outlet. The service areas are further divided into sub‐service areas. The number of households is per sub‐service area. The inset map shows the location of the study area within the Netherlands.
Figure 3
Figure 3
Workflow showing independent variables data generation process for training the SVM model. This is a sub‐step of Figure 1.
Figure 4
Figure 4
Effect of hotspot prevalence and sampling regime on SVM individual classification accuracy. The sampling regime shows the sampling frequency. For example, 1440 and 90 samples per day mean one sample every minute and 16 min, respectively.
Figure 5
Figure 5
(a) shows SVM individual accuracy versus time of measurement window for different window durations. (b) Shows the diurnal pattern for wastewater flow from toilet flushing.
Figure 6
Figure 6
Fast‐Fourier transform of concentration data, averaged for each of the four scenarios, 90 samples measured/day, 0.5% hotspot prevalence.
See this image and copyright information in PMC

References

    1. Abdeldayem, O. M. , Dabbish, A. M. , Habashy, M. M. , Mostafa, M. K. , Elhefnawy, M. , Amin, L. , et al. (2022). Viral outbreaks detection and surveillance using wastewater‐based epidemiology, viral air sampling, and machine learning techniques: A comprehensive review and outlook. Science of the Total Environment, 803, 149834. 10.1016/j.scitotenv.2021.149834 - DOI - PMC - PubMed
    1. Ahmed, W. , Angel, N. , Edson, J. , Bibby, K. , Bivins, A. , O'Brien, J. W. , et al. (2020). First confirmed detection of SARS‐CoV‐2 in untreated wastewater in Australia: A proof of concept for the wastewater surveillance of COVID‐19 in the community. Science of the Total Environment, 728, 138764. 10.1016/j.scitotenv.2020.138764 - DOI - PMC - PubMed
    1. Bailey, O. , Arnot, T. C. , Blokker, E. J. M. , Kapelan, Z. , Vreeburg, J. , & Hofman, J. (2019). Developing a stochastic sewer model to support sewer design under water conservation measures. Journal of Hydrology, 573, 908–917. 10.1016/j.jhydrol.2019.04.013 - DOI
    1. Banik, B. K. , Di Cristo, C. , Leopardi, A. , & de Marinis, G. (2017). Illicit intrusion characterization in sewer systems. Urban Water Journal, 14(4), 416–426. 10.1080/1573062X.2016.1176220 - DOI
    1. Bartos, M. , & Kerkez, B. (2021). Observability‐based sensor placement improves contaminant tracing in river networks. Water Resources Research, 57(7), e2020WR029551. 10.1029/2020WR029551 - DOI

LinkOut - more resources