ゼブラフィッシュ幼生における環境リスク曝露に対する炎症反応のAATE-UNet自動評価

In the evaluation of drugs/cosmetics toxicology/efficacy on livings, rapid assessment of inflammatory responses in zebrafish models is critical but hindered by labor-intensive manual neutrophil counting. To be addressed, this study developed the innovative AATE-UNet, a deep learning model that automates high-throughput image analysis for precise inflammation quantification. In result, this UNet-based architecture processes lateral zebrafish images to segment complex anatomical regions (yolk sac, spinal cord) and quantify neutrophils with 90 % accuracy (<10 % error versus manual counts), while slashing processing time from ∼1 h to < 5 min per sample. The advancement eliminates subjective variability and workflow bottlenecks inherent to manual methods. As a supplement, qPCR analysis revealed pollutant-driven dysregulation of inflammatory cytokines (e.g. IL-1β, IL-6, IL-10, and TNF-α), bridging cellular neutrophil dynamics with molecular pathways. Moreover, the AATE-UNet model was packaged as a user-friendly executable file (.exe), facilitating application in standard fluorescence imaging systems by enabling use on computers with compatible hardware without the need for specialized software or training. By correlating neutrophil thresholds with pollutant concentrations, our framework establishes actionable metrics for toxicity evaluation while offering a scalable solution to accelerate environmental risk assessments. Consequently, this study provides a reliable method effectively enables objective morphometric analysis of zebrafish larvae to evaluate the inflammatory responses of environmental exposures.