HumBug

Motivation

A vector is an organism (normally an insect or other arthropod) that transmits a disease agent (such as a virus, bacteria or parasite) from one host to another. Due to their blood feeding habits, female mosquitoes are particularly good at being vectors and mosquito-transmitted diseases are responsible for an estimated one million deaths each year.
(source: http://www.scientistsagainstmalaria.net)

Although there are approximately 3500 mosquito species in the world, not all are disease vectors. More than 70 species of the Anopheles genus are human malaria vectors able to spread human malaria (with approximately 40 species able to do so at a level of specific concern for human health). However, many closely related species look identical but due to their behaviour, one may be a highly dangerous vector whereas the other is harmless. Therefore the abundance and identity of mosquitoes in countries where vector-borne diseases are found need to be regularly monitored.

Control programmes, such as those distributing insecticide treated bed nets, have lead to significant reductions in malaria transmission. Indeed, recent estimates suggest that incident of clinical disease in Africa had fallen by 40% in just the last 5 years. Yet mosquito surveys are expensive and time consuming and they also put the person sampling the mosquitoes at risk of catching diseases. Furthermore, once a mosquito survey has been completed, it requires vector experts and costly analyses to distinguish one species (and possible vector) from the other. With emerging insecticide resistance and the wide distribution of species that naturally avoid indoor based interventions, current vector control regimes may begin to struggle. Their reliance on mapping and modelling tools based on sparse, poorly distributed static data may no longer be enough to deal with the residual and more resilient transmission nor the more complex, multi-parasite species found out of Africa.

Our methods

The shape, speed and movement of a mosquito’s wings as it flies causes a distinctive buzz, easily identifiable when heard in a bedroom at night. Analysing this sound (the flight tone) combined with additional information such as where and when the mosquito was flying, allows many mosquitoes to be identified.

Utilising portable devices, such as mobile phones, wristbands or other acoustic monitoring devices, we are developing a real-time detection system that can alert users to the presence of a vector species. To improve the machine learning algorithms that we have developed to detect and classify mosquitoes, we need huge amounts of flight tone data to train and refine it. When we record a flying mosquito, the sound of its beating wings is relatively quiet and can be lost within any background noise. As such, a recording of a flying mosquito that lasts an hour may only have a few seconds of audio when the mosquito flies close enough to the microphone to provide acoustic data loud enough for us to analyse. It is these snippets of sound we are trying to find – with your help. This huge task is impossible without citizen scientists' help. Thank you!