Multi-channel algorithm for segmentation of tumor blood vessels using multiplexed image data

Angiogenesis is the development of new vasculature from existing vasculature. The developmental biology and pathology of angiogenesis are major focuses of biomedical research. While many segmentation algorithms have been implemented for blood vessel analysis, they are limited by maturation dependent variation in blood vessel protein expression, expression by other cell types and discontinuous vessel staining due to tissue sectioning. We describe a method that combines image data from multiple blood vessel-expressed proteins to dynamically estimate optimal correlation metrics to drive improved vessel segmentation. A single marker approach resulted in overestimation of vessel number, whereas the multi-channel algorithm agreed closely with manual counts generated by two independent observers. Our results suggest that this new approach improves the vessel segmentation, which will be useful in the study of angiogenesis in health and disease.