Background
The age distribution of a mosquito population is a major determinant of its vectorial capacity. To contribute to disease transmission, a competent mosquito vector, carrying a pathogen, must live longer than the extrinsic incubation period of that pathogen to enable transmission to a new host. As such, determining the age of female mosquitoes is of significant interest for vector-borne diseases surveillance and control programs.
Methods
In this contribution, an automated age-grading system was developed to classify the age of female Culex pipiens, which is the primary vector of West Nile virus and other pathogens of medical and veterinary importance in northern latitudes. The system comprises an optical wingbeat sensor coupled to the entrance of a mosquito trap and a machine learning model. Three age classes were used in the study: young (2–4 days), middle (7–9 days) and old (14–16 days). From a balanced dataset of flight data, four features were extracted: wingbeat fundamental frequency, spectrogram, power spectral density and Mel frequency cepstral coefficients. The features were used for training with the XGBoost algorithm to generate a model for age classification.
Results
The best performing model was trained with the power spectral density feature on two age classes, ≤ 4 days old and ≥ 7 days old, and had an accuracy of 71.8%.
Conclusions
An automated mosquito age-grading system was applied for the first time to our knowledge for automated age classification in mosquitoes; and complements the mosquito genus and sex classification capability of the system reported in our previous work. The system may find use in mosquito-borne disease surveillance and control to help to discriminate young mosquitoes (≤ 4 days old) from older mosquitoes, which may act as vectors of arboviruses.
Graphical Abstract