Image-based classification of abamectin and fipronil exposure in Heteracris littoralis (Orthoptera: Acrididae) embryos using compound eye histology
Pesticide exposure is usually assessed by chemical analysis of residues, which says little about biological effect. We asked whether the compound eye of grasshopper embryos carries image features that reflect exposure. Embryos of Heteracris littoralis were treated with sublethal concentrations (LC30) of abamectin or fipronil, and ten malformed embryos per treatment and ten untreated controls were sectioned, stained with hematoxylin and eosin and photographed. The compound eye was extracted by graph-cut segmentation, thresholded, sharpened and rotated to a common orientation. Because only 30 embryos were available, each color channel was augmented by bit flipping, giving 1,440 channel images. Histogram of oriented gradients features were used to train five classifiers to assign images to control, abamectin or fipronil groups, with training and test sets drawn from different embryos and the split repeated ten times. Linear support vector machine and one-nearest-neighbor classifiers made no errors, logistic regression and a neural network reached 99.88% and 99.97% test accuracy, and decision trees 97.10%. These results show that the three exposure groups are separable in this small dataset. Validation on independently sampled embryos is needed before any wider use.
Authors
- Khaled Mohamed Abdel Rahman (ORCID: https://orcid.org/0000-0002-7009-9013)
- H. A. Ayad
- M. A. Rushdi
- M. H. Annaby
Institutions
- Cairo University (EG)
Publication Details
- Journal
- Research Square
- Published
- 2026-11-19
- DOI
- https://doi.org/10.21203/rs.3.rs-11550796/v1
- Primary Topic
- Environmental Toxicology and Ecotoxicology
- Type
- preprint