# Rapid and Nondestructive Sex Differentiation of Aedes aegypti Using Infrared Spectroscopy and Discriminant Analysis

## TL;DR

This paper introduces a fast and non-invasive method to distinguish male and female Aedes aegypti mosquitoes using infrared spectroscopy and machine learning, improving disease surveillance.

## Contribution

A novel nondestructive method for sex differentiation of Aedes aegypti using NIR and MIR spectroscopy with chemometric models is proposed.

## Key findings

- NIR and MIR spectroscopy achieved 93.94% sensitivity and specificity in sex differentiation.
- NIR models outperformed MIR models, with NIR-iSPA-PLS-DA showing the best classification results.
- The method is more cost-effective and faster than traditional techniques.

## Abstract

Aedes
aegypti (A.
aegypti) mosquitoes are arthropods that transmit a
set of viruses of great relevance to public health, such as Dengue,
Zika, and Chikungunya, causing recurrent epidemics in tropical and
subtropical regions. This fact raises the need for improvements and
enhancements in the entomological surveillance of these species, preventing
the occurrence of outbreaks. In this work, the potential of near-infrared
(NIR) and mid-infrared (MIR) spectroscopy together with variable selection
and supervised classification methods for the differentiation of A. aegypti mosquitoes according to their sex was
analyzed. 66 mosquitoes were analyzed by NIR and 198 by MIR. Chemometric
models of successive projection algorithm–linear discriminant
analysis (SPA-LDA), genetic algorithm–linear discriminant analysis
(GA-LDA), and partial least squares–discriminant analysis (PLS-DA)
coupled with interval selection by the successive projections algorithm
(iSPA-PLS-DA) were developed to build models capable of differentiating
the spectra between males and females. The results demonstrated that
sensitivity and specificity values achieved 93.94%, demonstrating
that it is possible to perform the differentiation with both NIR and
MIR spectroscopies. Based on sensitivity and specificity values, it
was possible to note a better performance of the NIR models compared
to the MIR models, with the NIR-GA-LDA model being more efficient
than the NIR-SPA-LDA model, while NIR-iSPA-PLS-DA was highlighted
with the best classification results. Furthermore, a cost and time
analysis against gold standard techniques is also performed in this
study, demonstrating the advantage of the tools described here. Finally,
new studies should be developed so that in the near future it will
be possible to count on portable tools based on spectroscopy in the
field of entomological surveillance of these species, seeking to identify
regions with a high probability of outbreaks in the future.