EpiVECS: exploring spatiotemporal epidemiological data using cluster embedding and interactive visualization

Abstract

The analysis of data over space and time is a core part of descriptive epidemiology, but the complexity of spatiotemporal data makes this challenging. There is a need for methods that simplify the exploration of such data for tasks such as surveillance and hypothesis generation. In this paper, we use combined clustering and dimensionality reduction methods (hereafter referred to as ‘cluster embedding’ methods) to spatially visualize patterns in epidemiological time-series data. We compare several cluster embedding techniques to see which performs best along a variety of internal cluster validation metrics. We find that methods based on k-means clustering generally perform better than self-organizing maps on real world epidemiological data, with some minor exceptions. We also introduce EpiVECS, a tool which allows the user to perform cluster embedding and explore the results using interactive visualization. EpiVECS is available as a privacy preserving, in-browser open source web application at https://episphere.github.io/epivecs.

Visualization of Spatial–Temporal Epidemiological Data: A Scoping Review

“…Task Accuracy User Experience Task Efficiency N/A [14,16,18,22,24,[27][28][29][30][32][33][34][35][37][38][39][41][42][43][44][45][46][47]49,[52][53][54][55][56][57][58][59]61,64,65,[68][69][70]72,73,[75][76][77][78][79][81][82][83][84][85][86][87]…”

“…The studies that used a map-style representation (79/114) or choropleth maps [17,23,24,27,30,32,34,35,37,40,46,48,49,52,55,56,58,60,62,64,68,71,[76][77][78][79]82,84,87,99,103,105,111,[113][114][115][116][119][120][121]124,127]. Choropleth maps are widely used to visualize the state of a variable across a geographic area.…”

“…Choropleth maps are widely used to visualize the state of a variable across a geographic area. In the papers, this technique was mostly used to represent the number of cases or cumulative cases in an area [17,22,24,27,46,48,52,56,58,[60][61][62][63]65,67,71,76,79,84,87,99,103,105,[113][114][115][116]119,121,127], along with the incidence of a disease [30,32,56,62,63,105,119,120,127]. It was also used to a lesser extent to visualize other measures, such as prevalence [35,64,76] and morbidity [111].…”

“…The studies that used a map-style representation (79/114) adopted different strategies: static geographic 2D or 3D maps [14,15,19,20,22,31,33,36,38,[42][43][44]50,75,81,83,88,93,102,104,107,109,117,118,123], dynamic 2D maps [20][21][22]25,61,65,80,85,89,90,94,97], or choropleth maps [17,23,24,27,30,32,34,35,37,40,46,48,49,52,…”

“…According to our review, more than half of the 114 studies (67/114) used 2D or 3D line charts as depictions of time series to represent temporal data related to a pan-demic. They were not only used to depict the number of cases per week, month, or year [16,21,22,24,25,30,36,50,53,54,56,59,60,62,65,67,68,[71][72][73]75,79,[84][85][86][87]97,98,100,103,104,106,109,113,114,120,121,127] or cumulative cases [22,27,52,55,65,70,71,[98][99][100]105,…”