Tracking of Cell Population from Time Lapse and End Point Confocal
Microscopy Images with Multiple Hypothesis Kalman Smoothing Filters
Angiogenesis is the formation of new blood vessels from a monolayer of cells or by the reorganization of capillaries via morphogenesis. When exposed to growth factors, endothelial cells forming a monolayer, undergo stochastic phenotype transitions such as migrating, quiescent, proliferation or death. Explorations of the angiogenic sprouting mechanism to determine how a population of cells could sprout out creating a new vascular network structures requires efficient and accurate image analysis to provide estimates of the cell trajectories and phenotypes over time.
A number of issues are faced in automating cell tracking in addition to the extensive volume of the experimental data. These issues include the low signal to noise ratio of the data and varying cell densities with a denser group of cells forming the monolayer and a more sparely populated cells migrating from the monolayer. As a result complex cellular topologies are encountered due to close contact and partial overlap of cells and the ability of the cell to deform and alter its shape. Cell densities also differ due to the proliferation and death of cells. Another issue specific to fluorescent confocal microscopy is photo-bleaching. As more images are acquired over time, more photochemical destructions occurs in the stained fluorescent molecules in the cells. Excessive light exposure stimulating these fluorescent molecules may even cause photo-toxicity. As a results, the sampling time and the acquisition step size throughout the height of the cell migration regions is limited.
Prof.Harry Asada's group at SMART BioSyM along with collaborator Assoc. Prof. Dr.Ang Marcelo of NUS, approach these multi-cell tracking challenges via probabilistic methodologies. A Kalman filtering combined with Multiple Hypothesis Testing (MHT) and smoothing/retrodiction is proposed to allow tracking of varying cell dynamics and account for clutter due to close contact cells. In addition to that, probabilistic techniques are used to incorporate fixed end-point imaging data with time-lapse information in a mathematically consistent manner.
The details were presented by Dr.Sharon Ong, a post-doctoral researcher @ SMART-BioSyM, in a paper "Tracking of Cell Population from Time Lapse and End Point Confocal Microscopy Images with Multiple Hypothesis Kalman Smoothing Filters" at the IEEE Computer Society Workshop on Mathematical Methods in Biomedical Image Analysis 2010, San Francisco, CA, USA, June-2010.