Conference Agenda
Overview and details of the sessions of this conference. Please select a date or location to show only sessions at that day or location. Please select a single session for detailed view (with abstracts and downloads if available).
|
Daily Overview |
| Session | |
|
SES 3-3-2: Model applications and development 2 Location: HSB1 Session Chair: Martin Fencl Session Chair: Jeroen Langeveld | |
    | |
1:30pm - 1:45pm
Deep learning methods for city-agnostic public health forecasts from wastewater-based epidemiological data 1: Université Laval, département de génie civil et de génie des eaux; 2: Thales Group; 3: NQB.ai,; 4: Université Laval, Département d’informatique et de génie logiciel Wastewater-Based Epidemiology (WBE) has been adopted as a low-cost, unbiased method of monitoring the spread of COVID-19. Viral signals can precede clinical testing by several days, making it an appealing basis for developing decision-support tools for public health interventions. However, WBE signals are subject to in-sewer processes that complicate their interpretation. Moreover, sewer dynamics vary between sewersheds, making the joint interpretation of WBE data between cities challenging. Wastewater quality data have been shown to help normalise WBE; however, it is unclear which wastewater characteristic can best perform this task. This study investigates the applicability of convolutional neural networks for developing short-term (7-day) forecasting models for public health indicators based on a WBE signal and wastewater quality data. A novel model structure is proposed to create a city-agnostic model that can be easily applied to new cities after it has been trained. The proposed model is tested on field data from 9 mid-sized North American cities. The model structure is found to perform better than simple models, warranting further investigation. Analysis of the model performance indicates that WBE signals support the prediction of new cases and that providing wastewater characteristics allows the model to better generalize its learning to new sewersheds. | |