Utilizing supervised machine learning to identify microglia and astrocytes in situ: implications for large-scale...

Area of Science:

• Histopathology
• Preclinical Research
• Machine Learning in Biology

Background:

• Histological tissue sample evaluation is vital for understanding disease and injury mechanisms.
• Manual analysis of high-resolution histological images is a time-consuming bottleneck in data acquisition.

Purpose of the Study:

• To describe and validate a novel machine learning-based analytical pipeline for detailed image analysis of cellular markers in histological samples.
• To demonstrate the pipeline's adaptability for multiple cell types and epitopes across different animal models.

Main Methods:

• Utilized the Opera High-Content Screening system and Harmony software for a machine learning-based analytical pipeline.
• Validated the pipeline using mouse and rat brain sections, analyzing single proteins (Iba1, GFAP) and co-localization (aquaporin-4, tomato lectin).

Main Results:

• The automated pipeline showed no significant differences compared to manual analysis for microglial and astrocyte markers (Iba1, GFAP).
• Efficient and accurate quantification of co-localized astrocytic endfeet on blood vessels was achieved.
• The automated platform completed analyses 200 times faster than manual methods while maintaining sensitivity and accuracy.

Conclusions:

• The automated high-throughput machine learning approach offers significant benefits for in situ tissue sample analysis.
• The pipeline is effective across different animal models, reducing analysis time and increasing research productivity.