Cell Segmentation for Division Rate Estimation in Computerized Video Time-Lapse Microscopy

摘要

The automated estimation of cell division rate plays an important role in the evaluation of a gene function in high throughput biomedical research. Using Computerized Video Time-Lapse (CVTL) microcopy , it is possible to follow a large number of cells in their physiological conditions for several generations. However analysis of this large volume data is complicated due to cell to cell contacts in a high density population. We approach this problem by segmenting out cells or cell clusters through a learning method. The feature of a pixel is represented by the intensity and gradient information in a small surrounding sub-window. Curve evolution techniques are used to accurately find the cell or cell cluster boundary. With the assumption that the average cell size is the same in each frame, we can use the cell area to estimate the cell division rate. Our segmentation results are compared to manually-defined ground truth. Both recall and precision measures for segmentation accuracy are above 95%.