Harald Birkholz (GRK Welisa)
from 09:00 to 10:30
|Where||Ulmenstr. 69, House 1, R: 134|
|Contact Name||Harald Birkholz|
|Contact Phone||+49 (0)381 498 6633|
|Add event to calendar||
Automatic Quantification of Cell-Morphological Parameters in Microscopic Images - Actin Filaments in Confocal Microscopic Images as Bright Ridge Network
This talk gives an overview of novel techniques for quantification of cell-morphological parameters from confocal laser scanning microscopic images. Human osteoblastic cells, among others, expose fibrous cytoskeletal structures or protrusions which are appropriately modelled as /bright ridges in digital images/. The /actin-filament/ as part of the cytoskeleton is relevant for assessing the response of cells to mechanical and chemical influences, e.g. induced by the contact to biomaterials on implant surfaces. An automated evaluation of parameters describing the filament will allow for in-depth research about the /cell-material-dialogue/.
The first part of the presentation is introductory. It is followed by a review of literature and the proposal of a novel ridge-tracking technique. The technique is based on a predictor-corrector concept and introduces context information more selectively than /scale-space/ and less sensitively than /medial-axis-transform techniques/. For stable results independent of image acquisition errors there is a need for preprocessing. The morphological /top-hat transform/ and a /total-variation-type method/ turn out to be suitable for the reduction of perturbations while preserving bright ridges. The subsequent parts of the talk are devoted to a set of parameters describing the detected fibres quantitatively concluded with an evaluation of accuracy. This statistical evaluation makes use of a /random-model/ of the data which is still in the process of development. So far this allows for a partial /verification of the image processing technique/. Given interest and some time left, there can be an additional demonstration of a /software-solution/ implementing the proposed image-processing and quantifications routines.