CenSegNet: a generalist high-throughput deep learning framework for centrosome phenotyping at spatial and single-cell...

Area of Science:

• Oncology
• Biotechnology
• Computational Biology

Background:

• Centrosome abnormalities (CA) are prevalent in epithelial cancers, but their complexity and heterogeneity are poorly understood.
• Conventional image analysis methods have limitations in accurately quantifying CA and their spatial context.
• Understanding CA is crucial for cancer diagnosis, prognosis, and therapeutic development.

Purpose of the Study:

• To develop and validate CenSegNet, a deep learning framework for high-throughput segmentation of centrosomes and epithelial architecture.
• To enable accurate, generalisable, and spatially resolved centrosome phenotyping across diverse imaging modalities and tissue types.
• To investigate the mechanistic uncoupling, spatial distribution, and clinical significance of numerical and structural CA in breast cancer.

Main Methods:

• Development of CenSegNet, a modular deep learning framework for automated segmentation of centrosomes and epithelial structures.
• Application of CenSegNet to a large cohort of breast cancer tissue microarrays (911 cores, 127 patients).
• Spatially resolved quantification of numerical and structural CA, analyzing their associations with clinical and genomic data.

Main Results:

• CenSegNet accurately segments centrosomes and epithelial architecture, enabling large-scale, spatially resolved CA quantification.
• Numerical and structural CA subtypes are mechanistically distinct, with unique spatial distributions and age-dependent dynamics.
• Structural CA correlate with overall survival, while discordant CA at tumour margins indicate local aggressiveness and stromal remodelling.

Conclusions:

• CenSegNet provides a scalable platform for detailed centrosome phenotyping in cancer.
• Spatially resolved CA analysis reveals distinct roles in breast cancer progression, survival, and tumour microenvironment interactions.
• This approach facilitates systematic investigation of centrosome biology in cancer and other epithelial diseases.