Deep Learning-Based Segmentation of Morphologically Distinct Rat Hippocampal Reactive Astrocytes After Trimethyltin...

Area of Science:

• Neuroscience
• Computational Biology
• Toxicology

Background:

• Astrocytes are key regulators of homeostasis in the central nervous system (CNS).
• Astrocytes undergo morphological changes following injury, a process known as reactive astrogliosis.
• Trimethyltin (TMT) is a neurotoxicant that induces reactive astrogliosis in rats, traditionally assessed via manual counting or scoring.

Purpose of the Study:

• To develop and validate a deep learning algorithm for quantitative assessment of reactive astrogliosis.
• To explore the potential of AI-driven image analysis in neurotoxicity studies.

Main Methods:

• A supervised deep learning algorithm was trained on 940 manually annotated astrocytes from rat hippocampus and cortex.
• The algorithm was applied to glial fibrillary acidic protein-labeled brain sections from a rat TMT model.
• The algorithm quantified individual astrocyte cell counts, areas, and circumferences.

Main Results:

• The deep learning algorithm accurately identified astrocytes of varying sizes with high confidence.
• A novel morphometric marker, derived from astrocyte cell area and circumference, was identified.
• This marker demonstrated a correlation with the time-dependent progression of TMT neurotoxicity.

Conclusions:

• Deep learning offers a powerful, quantitative approach for analyzing reactive astrogliosis in neurotoxicity studies.
• The developed astrocyte algorithm provides a reliable tool for high-throughput analysis of astrocyte morphology.
• This study underscores the potential of AI in advancing neuropharmacology and toxicology research.