Skip to main content
AMCA25 Main banner
Final Program


Icon Legend
This session is not in your schedule.
This session is in your schedule. Click again to remove it.
Back

New Vector Control Tools: UAS and Artificial Intelligance

New Vector Control Tools: UAS and Artificial Intelligance

289 - Improving IDX artificial intelligence vector identification tool for the identification of cryptic species: Culex pipiens pipiens and Culex quinquefasciatus

Thursday, March 6, 2025
1:55 PM – 2:05 PM AST
Location: 202
Abstract: Controlling vector-borne diseases like West Nile virus (WNV) relies on accurate mosquito species identification, which is often challenging due to morphological similarities. Convolutional neural networks (CNNs) for image recognition, a deep learning method, have emerged as a promising modality with the capability to visually differentiate between mosquito species, with emerging applications approaching early operational use cases on wild field-collected specimens. IDX, Vectech’s system for mosquito and tick imaging with integrated artificial intelligence identification, has been deployed to 51 organizations, some of whom are in a data gathering and validation phase, and others who are in an operational phase. IDX currently identifies 57 mosquito “species” globally, including species complexes or groups that appear morphologically similar. This includes Culex pipiens pipiens and Cx. quinquefasciatus, two species with different vectorial capacities and ecologies. Accurately differentiating these species is critical, as their overlapping ranges and similarities complicate surveillance and control. However, there have yet to be reliable keys for distinguishing between them with external morphology alone. Here we present preliminary findings in distinguishing between Cx. p. pipiens, Cx. quinquefasciatus, and their hybrids using CNN algorithms applied to IDX images. The true identification was determined using the Ace2 gene. Specimens were submitted by collaborators throughout the US and Canada, from locations representing northern (pipiens), southern (quinquefasciatus), and the hybridization zones. Specimens were imaged in IDX, followed by molecular genotyping. Using a dataset of 421 specimens for algorithm training and 37 specimens in testing, we achieved 94.6% and 90.2% micro and macro average recall respectively, indicating the feasibility of differentiating between these two species using our custom-built computer-vision algorithms on images taken on the IDX platform. This work is ongoing, pending continued data collection to expand the geographical coverage and validate the generalizability of the identification accuracy.