Abstract
Identifying areas in a landscape where individuals have a higher likelihood of disease infection is key to managing diseases. Unlike conventional methods relying on ecological assumptions, we perform a novel epidemiological tomography for the estimation of landscape propensity to disease infection, using GPS animal tracks in a manner analogous to tomographic techniques in positron emission tomography (PET). Treating tracking data as random Radon transforms, we analyze Cervid movements in a game preserve, paired with antibody levels for epizootic hemorrhagic disease virus (EHDV) -- a vector-borne disease transmitted by biting midges. After discretizing the field and building the regression matrix of the time spent by each deer (row) at each point of the lattice (column), we model the binary response (infected or not) as a binomial linear inverse problem where spatial coherence is enforced with a total variation regularization. The smoothness of the reconstructed propensity map is selected by the quantile universal threshold. To address limitations of small sample sizes and evaluate significance of our estimates, we quantify uncertainty using a bootstrap-based data augmentation procedure. Our method outperforms alternative ones when using simulated and real data. This tomographic framework is novel, with no established statistical methods tailored for such data.
Introduction
Predicting the spatial spread of infectious diseases across a landscape is a critical challenge for wildlife management and public health. Traditional approaches typically rely on "presence-only" data—recording where sick animals are found—which fails to account for where animals didn't go. Alternatively, complex ecological models are used, but these require prior knowledge of environmental drivers like temperature or vegetation.
Researchers have now developed a "tomographic" approach to mapping disease risk by treating GPS-tracked animal movements like medical imaging rays. By combining movement paths with infection status, this model can pinpoint disease hotspots—such as specific feeders or water sources—without any prior knowledge of the landscape’s physical features.
Epidemiological Tomography: The PET Scan Analogy
The core innovation of this work is the conceptual leap of treating animal movement as a sensor. In Positron Emission Tomography (PET), gamma rays travel through a medium, and their attenuation provides information about the density of that medium. In this epidemiological framework, a deer’s random path acts as the "ray," and the landscape’s "infection propensity" is the medium. If a deer becomes infected, it serves as a signal that its path likely intersected a high-risk zone.
Mathematical Framework
The researchers treat the mapping of disease risk as an inverse problem. The process involves several key steps:
- Landscape Discretization: The landscape is divided into a grid (lattice).
- Regression Matrix: A matrix is constructed where each row represents an individual animal and each column represents a grid cell. The values in the matrix correspond to the total time that animal spent in that specific cell.
- Total Variation (TV) Regularization: To ensure the resulting map is biologically plausible and not just a collection of noisy points, the model uses TV regularization. This encourages the identification of "piecewise constant" regions—distinct "blobs" of high or low risk that correspond to physical features like ponds or feeders.
- Quantile Universal Threshold (QUT): To avoid "detecting" false hotspots in small datasets, QUT is used to select the optimal level of smoothness, providing a robust way to control the false discovery rate.
Validation and Results
The model was tested using both synthetic landscapes (simulating features like lakes and rivers) and real-world data from a 172-hectare high-fenced ranch in Florida. The real-world study tracked 26 white-tailed deer and monitored their infection status for Epizootic Hemorrhagic Disease Virus (EHDV), a midge-borne illness.
- Superior Accuracy: In simulations, the tomographic TV method consistently outperformed standard smoothing techniques like Splines or Gaussian Process Regression.
- Real-World Success: In the Florida study, the model successfully pinpointed two specific feeders as high-risk zones for EHDV-1.
- Ecological Alignment: The reconstructed hotspots overlapped significantly with areas known to have high abundances of Culicoides stellifer (the biting midges that spread the virus), proving the model captured biological reality without any prior environmental input.
Conclusion and Future Directions
This framework opens the door for "medical imaging" of entire ecosystems. While the current model requires "before and after" blood samples (making it challenging for wild populations that aren't easily recaptured), it represents a significant step forward. Future work may involve "dynamic tomography" to track how disease hotspots shift seasonally. By treating movement as a primary data source rather than noise, researchers can now map the invisible landscape of disease risk with unprecedented precision.
Summary updated