Estimation of Cell Statistics on a Cell Manifold for High Content Screening with Automated Cell Segmentation

We present a method for estimating cell population statistics on a high dimensional nonlinear cell manifold constructed from the output of a high throughput cytometer that allows microscopic imaging of each cell. After active contour segmentation of the cell images, we construct a nonlinear cell manifold from the morphology characteristics and multispectral properties of the cells. We estimate the geodesic distance between the cells on the cell manifold and use it to estimate the cell population statistics on the manifold and quantify dissimilarities between cells to enable meaningful clustering of similar cell populations.