SATINN

SATINN automates classification and quantitative analysis of mouse testis histology by using convolutional neural networks to classify nuclei, stage seminiferous tubules, and extract cell- and tubule-level statistics from multiplexed immunofluorescence images for studies of development and pathology.

Key Features:

• Cell-type classification (CNNs): Convolutional neural networks classify nuclei within seminiferous tubules into seven distinct cell types with 94.2% accuracy.
• Tubule developmental staging (CNNs): A second convolutional neural network assigns seminiferous tubules to one of seven developmental stages with 90.4% accuracy.
• Multi-level framework: Two-stage analysis combines nucleus-level classification and tubule-level staging to provide hierarchical characterization of testis architecture.
• Derivation of statistical metrics: Extraction of numerous cell- and tubule-level statistics from wildtype mouse testis images for quantitative comparison.
• Application to pathological analysis: Enables quantitative description of developmental differences in mutant mouse lines and identification of deviations associated with disease or evolutionary changes.
• Integration with spatial technologies: Designed to interface conceptually with spatially-resolved genomic and proteomic technologies for combined histological and molecular analysis.

Scientific Applications:

• Testicular development and staging: Quantitative classification and staging of seminiferous tubules to study normal spermatogenic progression.
• Histopathology of the testis: Detection and description of pathological changes in mutant mouse lines and disease models.
• Developmental biology and evolution: Comparative analysis of developmental differences relevant to evolutionary studies.
• Integration with spatial omics: Correlating histological features with spatially-resolved genomic and proteomic data.

Methodology:

Convolutional neural networks analyze multiplexed immunofluorescence images from mouse testis: a first-stage CNN classifies nuclei into seven cell types, a second-stage CNN stages seminiferous tubules into seven developmental stages, and downstream routines extract cell- and tubule-level statistics.