Abstract
Image-based cell classification has become a common tool to identify phenotypic changes in cells. To date, these approaches are limited to model organisms with species-specific reagents available for cell phenotype identification, clustering and neural network training. Here we present Image3C (Image-Cytometry Cell Classification), a tool that enables cell clustering based on their intrinsic phenotypic features, combining image-based flowcytometry with cell cluster analysis and neural network integration. Using Image3C we recapitulated zebrafish hematopoietic cell lineages and identified cells with specific functions (phagocytes), whose abundance is comparable between treatments. To test Image3C versatility, we performed the same analyses on hemocytes of the snail Pomacea canaliculata obtaining results consistent with those collected by classical histochemical approaches. The convolutional neural network, then, uses Image3C clusters and image-based flowcytometry data to analyze large experimental datasets in an unsupervised high-throughput fashion. This tool will allow analysis of cell population composition on any species of interest.
Footnotes
↵# These authors share senior authorship
We've added a convolutional neural network that uses Image3C cluster information and Imagestream data for classification analysis of large datasets.