Statistical modeling and analysis of cell counts from multiplexed imaging data

The rapid development of multiplexed imaging technologies has enabled the spatial cartography of various healthy and tumor tissues. However, adequate statistical models are still lacking to compare tissue compositions across sample groups. Here, we developed two statistical models that accurately describe the distributions of cell counts in an imaging mass cytometry dataset comprising tissues from a lymph node, COVID-19-affected lung samples, and Hashimoto disease. The parameters of these distributions are directly linked to the field of view size and to cellular properties, including density and spatial aggregation. We identified statistical tests that improved statistical power for differential abundance testing compared with the commonly used rank-based test. Our analysis revealed spatial aggregation as the main determinant of statistical power and that high numbers of fields of view are required when cells are highly aggregated. To overcome this challenge, we propose a stratified sampling strategy that considerably reduces the required sample size.