A two-stage approach for segmenting spatial point patterns applied to multiplex imaging.

Area of Science:

• Computational biology
• Spatial statistics
• Immunohistochemistry

Background:

• Multiplex imaging allows cell localization within tissues.
• Tumor immunology research benefits from understanding spatial cell arrangements.
• Spatial patterns can indicate disease stages or treatment responses.

Purpose of the Study:

• To develop a method for partitioning multiplex tissue images based on spatial cell patterns.
• To identify distinct cellular regimes associated with disease or treatment.
• To apply spatial point pattern modeling for tumor immunology analysis.

Main Methods:

• A two-stage approach involving intensity and pair correlation function estimation.
• Dimensionality reduction of correlation functions using spectral decomposition.
• Bayesian hierarchical clustering with spatially-dependent labels and Markov Chain Monte Carlo sampling.

Main Results:

• Joint estimation and uncertainty quantification of cluster assignments and spatial characteristics.
• Demonstrated performance through simulations.
• Application to multiplex immunofluorescence images of pancreatic cancer tissue.

Conclusions:

• The proposed method effectively segments spatial point patterns into biologically relevant clusters.
• This approach provides a framework for analyzing complex spatial relationships in multiplex imaging data.
• The findings contribute to a deeper understanding of tumor microenvironments and treatment responses.