Novel cell segmentation and online learning algorithms for cell phase identification in automated time-lapse microscopy

Automated identification of cell cycle phases captured via fluorescent microscopy technique is very important for cell cycle understanding and drug discovery. In this paper, we propose a novel cell detection method that utilizes both the intensity and shape information of cell to improve the segmentation quality. In contrast to conventional off-line learning algorithms for classifcation, our study necessitates the on-line adaptivity to accommodate the ever-changing experimental conditions. An Online Support Vector Classifier (OSVC) is thus proposed, which features the removal of support vectors from the old model and assigning the new training examples with different weights according to their importance. Experimental results show the proposed system is effective for cell imaging segmentation and cell phase identification in time-lapse microscopy.