Biostatistics Journal Club: Quantifying Spatial Relationships Among Cells in Images
There is a burgeoning demand to analyze biomedical image data to quantify spatial relationships. Proximity of cell types to each other or to other features may provide clues to biological mechanisms, such as intercellular interactions. Standard point process regression models cannot easily be applied to such image data, in part because they lack physical coordinate systems. Jacqueline R. Starr, PhD, MS, MPH, of Brigham and Women’s Hospital has been adapting point process models to produce estimates of spatial clustering, i.e., the extent to which two cell types locate near (or away from) each other more (or less) than expected by chance. She has developed a meta-analytic framework to combine data across images and will present results from oral microbial biofilm image data to illustrate the method and types of spatial metrics that can be estimated with this multivariate approach.