Powerful computerized spatial epidemiology and semantics through the use of novel mathematical objects: Can artificial intelligence systems identify outbreak sources?

Abstract

We report on the use of artificial intelligence methods to identify the source of infectious disease outbreaks. The idea is to seek a probabilistic fit between data describing the problem being considered and a set of data providing the solution or to reconstruct "optimal data" given a specific set of rules or constraints. We used three examples to calculate both the Euclidean centroid using simple mathematics the hidden point using an evolutionary algorithm, and a new mathematical object: the topological weighted centroid. In the first (the 1854 London cholera epidemic) and second (the 1967 foot and mouth disease epidemic in England) examples the hidden point was within yards of the outbreak source. In the third example (the 2007 epidemic of Chikungunya fever in Italy) the hidden point was located in the river between the two village epicentres of the spread. Our results are consistent across examples and the method could provide an additional powerful tool for the investigation of the early stages of an epidemic. However, there is a need for field evaluation and validation of both methods and results.