Loading...
Using deep learning to assess the toxicological effects of sublethal exposure of a novel green pesticide in a stored‐product beetle
Casadei, Anita Boukouvala, Maria C Manduca, Gianluca Kavallieratos, Nickolas G Maggi, Filippo Ferrati, Marta Spinozzi, Eleonora Stefanini, Cesare DeSimone, Antonio Romano, Donato
Casadei, Anita
Boukouvala, Maria C
Manduca, Gianluca
Kavallieratos, Nickolas G
Maggi, Filippo
Ferrati, Marta
Spinozzi, Eleonora
Stefanini, Cesare
DeSimone, Antonio
Romano, Donato
Supervisor
Department
Robotics
Embargo End Date
Type
Journal article
Date
License
http://creativecommons.org/licenses/by/4.0/
Language
English
Collections
Research Projects
Organizational Units
Journal Issue
Abstract
Abstract BACKGROUND Managing stored‐grain pests requires new strategies to limit economic and health risks. This study analyses the sublethal effects of the natural compound carlina oxide on Prostephanus truncatus , providing new behavioural insights through a multidisciplinary approach. A fully automatic computer vision approach was developed to label two specific insect body parts, enabling the generation of an annotated dataset without manual intervention. This dataset was used to train a convolutional neural network (CNN) for pose estimation. A second dedicated CNN focused on the antennae to investigate neuroethological and sensory variations. RESULTS CNN for body parts detection achieved an average precision of 0.78, recall of 0.90, and F1 score of 0.84 on the test dataset. An additional CNN tracked key points for antennal pose estimation. Motor analysis showed that the LC 30 of carlina oxide reduced average speed and distance, induced altered exploratory behaviour, and affected thigmotaxis. Statistically significant features were evaluated using machine learning classifiers: random forest, support vector machine, and K‐nearest neighbours. The analysis comparing control and treated groups distinguishes LC 30 and LC 10 from the control group, while SHapley Additive exPlanation (SHAP) analysis explained the features contribution to predictions. CONCLUSIONS Metrics poorly distinguish individuals in the LC 10 and LC 30 classes, supporting the employment of lower sublethal concentration for the control of P. truncatus . However, our findings indicate possible neuroethological effects of green pesticides on sensory systems, highlighting the need for an accurate risk assessment to minimize ecosystem impacts and supporting integrated pest management within One‐Health and Eco‐Health frameworks. © 2026 The Author(s). Pest Management Science published by John Wiley & Sons Ltd on behalf of Society of Chemical Industry.
Citation
A. Casadei, M.C. Boukouvala, G. Manduca, N.G. Kavallieratos, F. Maggi, M. Ferrati, E. Spinozzi, C. Stefanini, A. DeSimone, D. Romano, "Using deep learning to assess the toxicological effects of sublethal exposure of a novel green pesticide in a stored‐product beetle," Pest Management Science, 2026, https://doi.org/10.1002/ps.70545.
Source
Pest Management Science
Conference
Keywords
31 Biological Sciences, 3109 Zoology
Subjects
Source
Publisher
Wiley