Deep Learning Assessment of Microplastic Toxicity in Freshwater Fish

Abstract

Microplastics have posed serious challenges to freshwater ecosystems owing to their persistent nature and negative impact on aquatic species. Microplastics affect fish in freshwater systems and cause physiological, behavioural, and biological impacts on them. Traditional toxicity tests for microplastics are time-consuming and inefficient at analysing complicated relationships between exposure of the environment and health responses of fish. The present study suggests developing a deep learning framework that assesses toxicity of microplastics in freshwater fish considering environmental exposure and biological response parameters. The suggested deep learning model integrates microplastic-related features such as exposure and particle features along with health responses like physiological and behavioural responses of fish. The deep learning model can effectively detect and classify toxicity of microplastics based on the patterns extracted from the combined data. The deep learning framework allows performing automatic analysis and interpreting biological response to microplastics. The suggested framework outperformed the performance of traditional machine learning models in terms of toxicity classification. The deep learning framework was able to accurately classify toxicity with 96.2% accuracy, 95.8% precision, 96.0% recall, and 95.9% F1-Score. The created deep learning-based assessment system offers an effective method to monitor freshwater pollution through automated toxicology and ecological risk assessments. The suggested system can facilitate the sustainable management of aquatic ecosystems and future AI-assisted environmental monitoring systems.