Skip navigation

Research

Information System for Medical Research and Education

The information system was primarily oriented to cerebrovascular diseases research. Its extension is possible because of a common design of an experiment-oriented system. It is possible to make an experiment (executable) with support for a user defined data evaluation, processing, etc. Data and meta data are stored together in a single database. An experiment could work only with a proper medical data files. Primarily, we use DASTA, DICOM, HL7 and re-use a structured data from the Safe Implementation of Treatments in Stroke (SITS) register. Requests are growing and this information system is also used for e.g. a liver research. A user friendly web interface is under the development, now. We also look for another collaborating centres.

In the future we plan e.g.:

De-identification of Medical Data

Our research uses only data without sensitive personal data. Removal of patients identity from used medical data perform consistently. We have created processes and methods to detect images of various modalities containing personal data. We have to remove (black out) all burned-in annotation of patients in the image before its use in any research.

Infarction Core Detection

Strokes are one the most frequent cause of death in population over 60 years. We focus on automatization of infarction core delineation using only non-contrast computed tomography and computed tomography angiography. We make perfusion blood volume maps by examinations subtraction. The final step - delineation itself - is our current challenge. We use techniques like thresholding and symmetry application with validation against follow-up examinations.

Infarction core delineation (white color), left column pictures are produced at admission time, right column are follow-up findings

Vascular Network Segmentation of The Liver Portal Vein

Geometry detection has been already solved but finding any suitable non-commercial software is not so easy. We propose universal procedure for geometry information detection. In order to liver's vascular network we want to find geometry of portal vein. Input examination used is computed tomography angiography stored in DICOM format.

The next step will be to find surface model of vascular network and describe it as coordinations of vertexes and list of connections between them forming triangles or rectangles lying on surface. The ideal surface should be formed by a mesh describing smooth tubes (cylinders) of diameters corresponding to detected veins.

Portal vein network