Method for organ segmentation

The inventions are serving to segment organs or organ areas out of multi-channeled volumetric MRT data sets.
To do so, the first invention uses segmentation outcomes from Trial data, which was generated by hand and recorded by several people. After recording an MRT dataset, a first pre-segmentation method is used.In this case, the grayscale information from the MRT data is used along with characteristic regions to create probability data for the organ.
A further development of this technique delivers the possibility to use the image slices from the training data to determine at least three classes of characteristic shape. For every class a support vector machine is trained. By using an exclusion method on each other they are able to detect the characteristic organ form. Based on this segmented area a patient specific probability data set is generated, on which a further segmentation technique is used. By combining both methods, a reliable fully automated segmentation is possible, with one or more emphases. Furthermore this application is less susceptible to variability in the MRT data. It is also possible to process one channel volumetric data.

Further information: PDF

PVA Mecklenburg-Vorpommern AG
Phone: +49 (0)381 49747-38

Contact
Christian Tholen

As Germany's association of technology- and patenttransfer agencies TechnologieAllianz e.V. is offering businesses access to the entire range of innovative research results of almost all German universities and numerous non-university research institutions. More than 2000 technology offers of 14 branches are beeing made accessable to businesses in order to assure your advance on the market. At www.technologieallianz.de a free, fast and non-bureaucratic access to all further offers of the German research landscape is offered to our members aiming to sucessfully transfer technologies.