Biology Reference
In-Depth Information
and Nutrition. The top graph depicts in gray time series of daily counts
of all reportable disease occurrences in the region on Colombo, Sri Lanka.
The alerts generated by the bi-variate temporal scan (shown as black dots)
executed in the retrospective screening mode, indicate days in late sum-
mer of 2008 when the increases in these counts exceeded the nationwide
trends. The analysts then run temporal scan screening against the data
from Colombo. It identified that patients diagnosed with leptospirosis are
the main contributing factor to the observed unusual trend (middle graph).
Even though the disease outbreak escalated slowly over more than 45 days,
the first meaningful alert from temporal scan was generated within three
days of its onset. Further drill-down revealed that the leptospirosis event
was not restricted to Colombo, but, as indicated by the Bayesian spatial scan
analysis, it has shortly spread to seven other regions. High probability of the
locations being affected by it is shown with large circles on the map in the
bottom graph. Bayesian spatial scan tracked probabilities of leptospirosis
outbreak anywhere in the nation for all days in the past data (plotted as the
upper line in the plot under the map). On August 20, 2008, that probabil-
ity exceeded 97%. Thanks to the efficiencies of T-Cube, the analysts could
detect, identify, and geographically localize leptospirosis event within a few
minutes of loading the data.
The RTBP system uses the T-Cube Web Interface (TCWI) as a platform to
visualize and manipulate multidimensional time series. Variants of TCWI
are also used elsewhere (Ray et al. 2008). Figure 9.3 presents a screenshot
of a pane of the interface showing an overlay of spatiotemporal distribu-
tions of the results of food tests against
Salmonella
collected by the U.S.
Department of Agriculture (USDA), and the counts of human cases of
salmonellosis recorded by the Centers of Disease Control and Prevention
(CDC). T-Cube Web Interface has enabled the USDA and CDC researchers
to browse their data jointly. It allowed them to drill into details, to focus on
specific
Salmonella
serotypes, to screen the data temporally and spatially as
well as with respect to other dimensions in it, and to do all that on-the-fly
in a fully interactive mode. This capability has led to discoveries of pat-
terns in data that can be indicative of food-borne sources of human disease
outbreaks. A generic variant of TCWI is also available for public evaluation
(Auton Lab 2009).
T-Cube has emerged as an enabling technology that supports automated
analyses as well as user manipulations of typical bio event data. By speeding
up execution of evidence aggregation procedures, it enables more compre-
hensive and timelier surveillance of such data with virtually no additional
burden on computing resources. It also enables closer-than-before interac-
tions between users and their data. Thanks to rapid responses to ad-hoc que-
ries, the analyst routine can be transformed from batch-mode processing of
data into much more interactive and exciting real-time navigation through it.
