Quantitative cell morphology featurization in multiplexed immunofluorescence images reveals tumor subtypes in cancer microenvironments

Cell morphology, the study of cellular form and structure, provides valuable insights into the diverse characteristics and functions of different cell types across various modalities. Cell morphology is also indicative of cellular activities and processes, such as cell division, motion, migration, differentiation, and responses to environmental cues. With advances in multiplexed immunofluorescence (mIF) imaging techniques, a powerful technique that allows spatial characterization of dozens of molecules at subcellular resolution, examining in situ cell morphology at scale is possible. Along with molecular information gleaned from mIF imaging, morphology features provide an additional axis of human-interpretable information that can help characterize tumor microenvironments and their component cells. Here, we describe a suite of morphology metrics and their applied utility in characterizing tumor microenvironments. We developed a suite of tools for cell morphology featurization and analysis to thoroughly understand the cells, their neighborhoods, and their interactions using cell morphology in multiplexed immunofluorescence images. The extracted morphology features include measurements of each single cell in four categories: orientation, area, length, and shape. These features are derived from cell segmentation, and use both cell intrinsic shape information, as well as contextual information about the relative orientations or alignment of cells. Using the new morphology features, we enhanced analysis methods for cell type phenotyping, cell morphology quantification, tissue partition, and clinical outcome predictions.