A massively parallel space-time connected approach based on implicit active contour methods to track leukocytes observed by multiphoton intra vital and confocal microscopy

Recent advances in biological imaging have allowed the development of powerful techniques such as multiphoton intravital microscopy or confocal microscopy that permits the imaging of immune cell interactions in vivo, opening the possibility to a new way of looking at cellular dynamics in the immune system. However, the analysis of the data generated by these techniques faces different problems mainly related with the application of proper algorithms to specifically track the migration patterns of the cells. Although there is a huge variety of approaches available in the literature, they can either not be applied in a "black-box" manner to specific problems, as many application specific parameters have to be tuned, or they require a substantial amount of time consuming manual support. Moreover, in the case of three-dimensional data obtained over many time frames, a large amount of data is obtained whose analysis can only be done with significant computational efforts. Thus, in order to improve the cell tracking analysis and to minimize the use of the manual tracking the available algorithms need to be improved to successfully predict cell behavior. It is therefore the main goal of this proposal, to develop and realize a cell-tracking approach, which allows on the one hand for the fast and reliable detection of cell movements and, on the other hand, is also capable of dealing efficiently with large data sets. This will be done within the specific framework of leukocytes observed by multiphoton intravital and confocal microscopy. By confining us to this application within this project, we ensure the applicability of the new methods and their implementation for ongoing and future research, as we can design them along our specific needs. It needs to be stressed, though, that due to the principal newness of our approach the new developments nevertheless can serve as a sound basis for future and broader applications. In extension of existing cell-tracking approaches, within this project we aim at exploiting the available information in space and time simultaneously. More precisely, instead of tracking the cells "step-by-step", we will develop and implement algorithms which will compute the trajectories of single cells as a whole, thereby improving the accuracy and reliability of the tracking process significantly. As an additional novelty of our approach, we furthermore aim at tracking algorithms, which are suitable for massively parallel data evaluation. The motivation for this is twofold: On the one hand, our coupled space-time approach requires larger amounts of data to be handled and processed simultaneously, which requires substantial computational efforts. On the other hand, high computational power is delivered nowadays by means of high concurrency, making the use of parallel algorithm inevitable. In order to achieve the above goals, obviously a highly interdisciplinary approach is required. Competences in biology, biological imaging, shape analysis, as well as in parallel computing and computational sciences are required. These different and interdisciplinary competences have been brought together with the broad and interdisciplinary applicant group of this proposal.