Deep Learning-Assisted identification and quantification of cell-associated microplastics using darkfield...

Area of Science:

• Environmental Science
• Toxicology
• Biomedical Imaging

Background:

• Microplastic (MP) interaction with human cells presents health risks.
• Quantifying MP-cell interactions and their extent is challenging.
• Existing methods lack single-cell resolution and high throughput.

Purpose of the Study:

• To develop an automated strategy for detecting and quantifying cell-associated microplastics at the single-cell level.
• To assess the dose-dependent effects of polystyrene microplastics on Caco-2 cells.
• To establish a robust method for microplastic toxicology studies.

Main Methods:

• Combined darkfield hyperspectral imaging (HSI) with a deep learning pipeline.
• Utilized Mask R-CNN for cell segmentation, LS-SVM for particle classification, and CHT for particle counting.
• Validated the pipeline's performance with high precision and accuracy metrics.

Main Results:

• The automated pipeline achieved high performance: 95% cell detection precision, 99.7% particle classification accuracy, and 99.6% particle detection precision.
• Demonstrated a dose-dependent relationship between PS MP concentration and uptake in Caco-2 cells.
• Observed no impact on cell viability at low concentrations, but significant viability reduction at high concentrations.

Conclusions:

• The integrated HSI and deep learning approach provides robust, single-cell resolution for quantifying MP-cell interactions.
• This method offers a valuable tool for microplastic toxicology, adaptable to various particle types and cell lines.
• The findings highlight the potential health risks associated with microplastic exposure at high concentrations.