Efficient AI-supported quantification of network structures in SEM imaged human mucus: A detailed protocol

Area of Science:

• Biomedical Engineering
• Microscopy
• Computational Biology

Background:

• Human mucus, found in the respiratory tract and ileum, possesses complex pore and filament structures crucial for its function.
• Manual analysis of these microstructures in scanning electron microscopy (SEM) images is time-consuming and prone to variability.

Purpose of the Study:

• To develop and validate an AI-assisted workflow for automated analysis of pore and filament structures in human mucus SEM images.
• To enhance the efficiency, precision, and reproducibility of mucus microstructure quantification.

Main Methods:

• A machine learning workflow was implemented in Fiji, utilizing a multi-layer perceptron classifier for pixel classification of SEM images.
• The workflow involved image preprocessing, classification, feature extraction, and data pooling for automated analysis.
• Manual and AI-assisted measurements were compared on respiratory and ileal mucus samples to validate the method.

Main Results:

• The AI-assisted workflow achieved high agreement with manual measurements, demonstrating superior efficiency and reduced data variability compared to human observers.
• The automated analysis showed robustness, as increasing the counting frame area did not significantly alter the resulting distributions.
• Effect sizes indicated that AI-generated data variability was lower than inter-observer variation in manual measurements.

Conclusions:

• The AI-assisted tool provides a precise, efficient, and high-throughput method for analyzing biological network structures in mucus.
• This technology can accelerate research into mucus alterations in diseases like cystic fibrosis and Crohn's disease.
• The workflow aids in understanding mucus pathophysiology and developing targeted therapeutic strategies.