Segmentation and Tracking of Confocal Images of Arabidopsis Thaliana Root Cells Using Automatically-Initialized Network Snakes

V Sethuraman, S Taylor, TP Pridmore, AP French & D Wells

The extraction of quantitative information from confocal microscope images continues to be a demanding and complex task. To this end, an automatic tool has been developed to support segmentation and tracking of networks of cells in the root of the model plant Arabidopsis thaliana. The approach is based upon automatically initialized Network Snakes. Network Snakes are a modification of the traditional ‘snakes’ technique, connecting several together to form a global topological structure. Global information helps in constraining object shape throughout the segmentation process, providing accurate boundary descriptions and allowing cell identity to be maintained over time. This technology is coupled with a novel automatic initialization method which modifies the watershed algorithm. The tool is demonstrated on a set of confocal microscope images showing root cells of Arabidopsis thaliana.