Statistical analysis of spatial patterns in tumor microenvironment images.

Abstract

Advances in tissue labeling, imaging, and automated cell identification now enable the visualization of immune cell types in human tumors. However, a framework for analyzing spatial patterns within the tumor microenvironment (TME) is still lacking. To address this, we develop Spatiopath, a null-hypothesis framework that distinguishes statistically significant immune cell associations from random distributions. Using embedding functions to map cell contours and tumor regions, Spatiopath extends Ripley's K function to analyze both cell-cell and cell-tumor interactions. We validate the method with synthetic simulations and apply it to multi-color images of lung tumor sections, revealing significant spatial patterns such as mast cells accumulating near T cells and the tumor epithelium. These patterns highlight differences in spatial organization, with mast cells clustering near the epithelium and T cells positioned farther away. Spatiopath enables a better understanding of immune responses and may help identify biomarkers for patient outcomes.