FINS: An interactive platform for automated zebrafish image analysis and morphological screening

Area of Science:

• Developmental Biology
• Toxicology
• Bioinformatics

Background:

• Zebrafish embryos are valuable models for developmental toxicity screening due to their transparency and conserved biology.
• Manual scoring of zebrafish embryos is time-consuming, variable, and difficult to scale for high-throughput needs.
• Lack of automated tools hinders the use of zebrafish assays in regulatory toxicology.

Purpose of the Study:

• To introduce FINS (Fish Imaging and Neural Scoring), a web-based platform for automated zebrafish morphological screening.
• To enable efficient and reproducible analysis of zebrafish embryos for developmental toxicity.
• To facilitate the adoption of zebrafish assays in large-scale research and regulatory applications.

Main Methods:

• Developed FINS, an interactive platform integrating deep learning (CNNs) for phenotype classification.
• Implemented intuitive visualization, manual correction, and quality control (QC) tools within FINS.
• Utilized multi-well plate images for automated identification of normal and abnormal zebrafish phenotypes.

Main Results:

• FINS demonstrated high classification accuracy and F1 scores for both normal and abnormal zebrafish phenotypes.
• The integrated QC module automatically identifies and flags unusable images, ensuring data quality.
• FINS reduced annotation time by over 70% compared to conventional manual scoring methods.

Conclusions:

• FINS provides a reproducible, scalable, and standardized solution for zebrafish image analysis.
• The platform's speed, accuracy, and interpretability support its use in research and regulatory settings.
• FINS facilitates the broader adoption of zebrafish embryo assays in toxicology pipelines.