MORPHIOUS

MORPHIOUS applies unsupervised machine learning to detect microglial and astrocytic activation from morphological features in brain tissue.

Key Features:

• Unsupervised learning: Employs a one-class support vector machine (SVM) in conjunction with density-based spatial clustering of applications with noise (DBSCAN) to learn reference morphologies of non-activated astrocytes and microglia and identify activation clusters.
• Morphology-based activation detection: Detects activation-associated morphological changes such as increased ionized calcium-binding adapter molecule 1 expression, enlarged soma size, and deramification without relying solely on cellular markers.
• Microglia classification: Distinguishes focal and proximal microglia based on their spatial relation to the activating stimulus.
• Astrocyte cluster identification: Identifies astrocyte clusters exhibiting changes in glial fibrillary acidic protein expression and branching.
• Validation via co-localization: Validates classifications by assessing co-localization with transforming growth factor beta 1 for focal and proximal microglia and with Nestin for proximal astrocytes.
• Application in neurodegeneration models: Correlates activation clusters with amyloid-β plaque load in hippocampal sections from TgCRND8 mice.

Scientific Applications:

• Brain injury and degeneration: Characterizes microglial and astrocytic responses in models of brain injury and neurodegeneration.
• FUS-induced BBB permeability studies: Detects activation triggered by blood-brain barrier (BBB) permeabilization using focused ultrasound (FUS).
• Amyloidosis models: Maps activation clusters and correlates them with amyloid-β plaque load in TgCRND8 mice.

Methodology:

Uses a one-class SVM to model non-activated glia morphologies and DBSCAN for spatial clustering to identify activation clusters; classifies microglia as focal or proximal based on spatial relation to the activating stimulus and assesses co-localization with transforming growth factor beta 1 and Nestin for validation.