GRAPHICALMODEL-BASEDTRACKINGOF.PDFVIP

GRAPHICAL MODEL-BASED TRACKING OF CURVILINEAR STRUCTURES IN BIO-IMAGE SEQUENCES Pradeep Koulgi, Mehmet Emre Sargin, Kenneth Rose and B. S. Manjunath ? Department of Electrical and Computer Engineering, University of California Santa Barbara, USA {pradeep, msargin, rose, manj}@ece.ucsb.edu Abstract tracks in the presence of high clutter from other micro- tubules in the vicinity, our algorithm is designed to over- Tracking of curvilinear structures is a task of funda- come the clutter and successfully track the contour. mental importance in the quantitative analysis of bio- Microtubules are ?lamentous subcellular structures logical structures such as neurons, blood vessels, reti- that are composed of tubulin protein sub-units. These nal interconnects, microtubules, etc. The state of the sub-units, by adding on to and dissociating from the art HMM-based contour tracking scheme for tracking tubulin polymer, make the microtubules grow and microtubules, while performing well in most scenarios, shorten over time. They are also capable of slight lateral can miss the track if, during its growth, it intersects an- motion over time, making them highly dynamic struc- other microtubule in its neighbourhood. In this paper tures. They are known to play a signi?cant role in many we present a graphical model-based tracking algorithm essential cellular processes [4]. Figure 1 shows a sam- which propagates across frames information about the ple image of microtubules. dynamics of all the microtubules. This allows the al- gorithm to faithfully differentiate the contour of interest from others that contribute to the clutter, and maintain tracking accuracy. We present results of experiments on real microtubule images captured using ?uorescence microscopy, and show that our proposed scheme out- performs the existing HMM-b