Integrating artificial intelligence and wing geometric morphometry to automate mosquito classification.

Autor: de Lima VR; Programa de Pós-Graduação em Medicina Tropical, Faculdade de Medicina, Instituto de Medicina Tropical, Universidade de São Paulo, São Paulo, SP 05403-000, Brazil., de Morais MCC; Instituto Israelita de Ensino e Pesquisa Albert Einstein (IIEPAE), Sociedade Beneficente Israelita Brasileira Albert Einstein (SBIBAE), São Paulo, SP, Brazil; Computational Systems Biology Laboratory (CSBL), Institut Pasteur de São Paulo, São Paulo, SP 05508-020, Brazil., Kirchgatter K; Programa de Pós-Graduação em Medicina Tropical, Faculdade de Medicina, Instituto de Medicina Tropical, Universidade de São Paulo, São Paulo, SP 05403-000, Brazil; Laboratório de Bioquímica e Biologia Molecular, Instituto Pasteur, São Paulo, SP 01027-000, Brazil. Electronic address: karink@usp.br.
Jazyk: angličtina
Zdroj: Acta tropica [Acta Trop] 2024 Jan; Vol. 249, pp. 107089. Date of Electronic Publication: 2023 Dec 01.
DOI: 10.1016/j.actatropica.2023.107089
Abstrakt: Mosquitoes (Diptera: Culicidae) comprise over 3500 global species, primarily in tropical regions, where the females act as disease vectors. Thus, identifying medically significant species is vital. In this context, Wing Geometric Morphometry (WGM) emerges as a precise and accessible method, excelling in species differentiation through mathematical approaches. Computational technologies and Artificial Intelligence (AI) promise to overcome WGM challenges, supporting mosquito identification. AI explores computers' thinking capacity, originating in the 1950s. Machine Learning (ML) arose in the 1980s as a subfield of AI, and deep Learning (DL) characterizes ML's subcategory, featuring hierarchical data processing layers. DL relies on data volume and layer adjustments. Over the past decade, AI demonstrated potential in mosquito identification. Various studies employed optical sensors, and Convolutional Neural Networks (CNNs) for mosquito identification, achieving average accuracy rates between 84 % and 93 %. Furthermore, larval Aedes identification reached accuracy rates of 92 % to 94 % using CNNs. DL models such as ResNet50 and VGG16 achieved up to 95 % accuracy in mosquito identification. Applying CNNs to georeference mosquito photos showed promising results. AI algorithms automated landmark detection in various insects' wings with repeatability rates exceeding 90 %. Companies have developed wing landmark detection algorithms, marking significant advancements in the field. In this review, we discuss how AI and WGM are being combined to identify mosquito species, offering benefits in monitoring and controlling mosquito populations.
Competing Interests: Declaration of Competing Interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
(Copyright © 2023 Elsevier B.V. All rights reserved.)
Databáze: MEDLINE
Externí odkaz: