Toxicity data for: COLLEMBOT: AI-based counting of Collembola for OECD 232 Tests

Résumé fourni par la source

This dataset is part of a scientific article, COLLEMBOT: AI-based counting of Collembola for OECD 232 Tests, in Environmental Toxicology & Chemistry: https://doi.org/10.1093/etojnl/vgag068. This dataset contains raw and processed data from laboratory toxicity tests on Folsomia candida (Collembola) exposed to various soil contaminants under controlled conditions. The study compares traditional manual counting of juveniles and adults with automated image-based counting using the COLLEMBOT system. Data were collected across multiple soils (LUFA 2.2, OECD variants) and exposure scenarios, including different concentrations of pesticides (e.g., imidacloprid, chlorpyrifos, lindane), fungicides (fluazinam, cyproconazole), microplastics (polystyrene), and reference substances (boric acid). The dataset includes: Raw observations: survival and reproduction endpoints per replicate. Manual vs automated counts: paired measurements for validation of automated image analysis. Metadata: compound identity, soil type, concentration (nominal and adjusted), replicate information. Dose-response modeling script: ECx values (ED10–ED90), NOEC/LOEC estimates, and statistical comparisons between counting methods. This resource supports ecotoxicological research, automation in soil toxicity testing, and reproducibility in environmental risk assessment workflows. There is an online version available at Huggingface, we are happy to recieve feedback on the usability: https://huggingface.co/spaces/MichaWehrli/COLLEMBOT_BETA_v0.1.0_under_development This repository also contains the model weights for COLLEMBOT under a AGPL 3.0 license with the code being available at GitHub under a MIT license: https://github.com/waldstrom/collembot

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.