UOC Neural Network Identifies Tiger Mosquitoes
July 21, 2021
study by researchers in the Scene understanding and artificial
intelligence (SUNAI) research group, of the Universitat Oberta de
Catalunya's (UOC) Faculty of Computer Science, Multimedia and
Telecommunications, has developed a method that can learn to identify
mosquitoes using a large number of images that volunteers took using
mobile phones and uploaded to Mosquito Alert platform.
Citizen science to investigate and control disease-transmitting
As well as being annoying because of their bites, mosquitoes can be the
carriers of pathogens. Rising temperatures worldwide are facilitating
their spread. This is the case with the tiger mosquito, Aedes albopictus,
and other species in Spain and around the world. As these species
spread, the science dedicated to combating the problems associated with
them develops. This is how Mosquito Alert was set up, a citizen science
project coordinated by the Centre for Research on Ecology and Forestry
Applications, the Blanes Centre for Advanced Studies and the Universitat
Pompeu Fabra to which UOC researchers have contributed.
This project brings together information collected by volunteer
citizens, who use their mobile phones to capture mosquito images as well
as that of their breeding sites in public spaces. Along with the photo
the location of the observation and other necessary information to help
in the identification of the species are also collected. This data is
then processed by entomologists and other experts to confirm the
presence of a potentially disease-carrying species and alert the
relevant authorities. In this way, with a simple photo and an app,
citizens can help to generate a map of the mosquitoes' distribution all
over the world and help to combat them.
"Mosquito Alert is a platform set up in 2014 to monitor and control
disease-carrying mosquitoes," says Gereziher Adhane, who worked on the
study with Mohammad Mahdi Dehshibi and David Masip. "Identifying the
mosquitoes is fundamental, as the diseases they transmit continue to be
a major public health issue. "The greatest challenge we encountered in
identifying the type of mosquito in this study was due to images taken
in uncontrolled conditions by citizens", he comments. He explains, the
image was not shot in close-up, and it contains additional objects,
which could reduce the performance of the proposed method. Even if the
images were taken up close, they were not necessarily at an angle that
entomologists could quickly identify, or because the images were taken
of killed mosquitos, the mosquito body patterns were deformed.
"Entomologists and experts can identify mosquitoes in the laboratory by
analysing the spectral wave forms of their wing beats, the DNA of larvae
and morphological parts of the body," Adhane points out. "This type of
analysis depends largely on human expertise and requires the
collaboration of professionals, is typically time-consuming, and is not
cost-effective because of the possible rapid propagation of invasive
species. Moreover, this way of studying populations of mosquitoes is not
easy to adapt to identify large groups with experiments carried out
outside the laboratory or with images obtained in uncontrolled
conditions," he adds. This is where neural networks can play a role as a
practical solution for controlling the spread of mosquitoes.
Deep neural networks, cutting-edge technology for identifying mosquitoes
Neural networks consist of a complex combination of interconnected
neurons. Information is entered at one end of the network and numerous
operations are performed until a result is obtained. A feature of neural
networks is that they can be trained through supervised,
semi-supervised, or unsupervised manner to process data and guide the
network about the type of result being sought. Another important
characteristic is their ability to process large amounts of data, such
as those submitted by volunteers participated in Mosquito Alert project.
The neural network can be trained to analyse images, among other data
types, and detect small variations that could be difficult for experts
to easily perceive.
"Manual inspection to identify the disease-carrying mosquitoes is
costly, requires a lot of time and is difficult in settings outside the
laboratory. Automated systems to identify mosquitoes could help
entomologists to monitor the spread of disease vectors with ease", the
UOC researcher emphasizes.
Conventional machine learning algorithms are not efficient enough for
big data analysis like the data available in Mosquito Alert platform,
because it contains many details and there is a high degree of
similarity between the morphological structures of different mosquito
species. However, in the study, the UOC researchers showed that deep
neural networks can be used to distinguish between the morphological
similarities of different species of mosquito, using the photographs
uploaded to the platform. "The neural network we have developed can
perform as well or nearly as well as a human expert and the algorithm is
sufficiently powerful to process massive amounts of images," says Adhane.
How does a deep neural network work?
a deep neural network receives input data, information patterns are
learned through convolution, pooling, and activation layers which
ultimately arrive at the output units to perform the classification
task," the researcher tells us, describing the complex process hidden
behind this model.
"For a neural network to learn there has to be some kind of feedback, to
reduce the difference between real values and those predicted by the
computing operation. The network is trained until the designers
determine that its performance is satisfactory. The model we have
developed could be used in practical applications with small
modifications to work with mobile apps," he explains. Although there is
still much development work to do the researcher concludes that "using
this trained network it is possible to make predictions about images of
mosquitoes taken using smartphones efficiently and in real time, as has
happened with the Mosquito Alert project."