SPARTA

© dkfz.de

50 % of cancer patients are treated with radiotherapy. The research project SPARTA (Software platform for adaptive multimodal radio- and particle therapy with autarkic expandability) has the goal to develop new software to support physicians in radiotherapy treatment planning. This software aims at a more efficient treatment that destroys only the tumor and protects the surrounding tissue. Together with collaborators within the DKFZ and other scientific and industrial partners we are working to integrate the newest scientific insights into the software.

Our specific goal at SIDT is to precisely quantify the already given radiation dose by tracking the changes in patient morphology. These changes can be formalized by registering the image data acquired at each treatment fraction to the acquired planning data. This allows the comparison of planned versus actual dose. Based on this information, we plan to predict the further course of irradiation and necessary changes for radiation treatment to achieve the goal, the extinction of the tumor. Currently, security margins are used in radiotherapy treatment planning to compensate for existing uncertainties, like positioning errors, organ movement, radiobiological or tumor shrinkage. Our goal is to quantify these uncertainties and evaluate their impact on radiation dose.

Further information:


Contact: Dörte Corr, Clemens Hentschke, Ralf O. Floca

Towards a generic semantically connected information model for medical image processing

Current research in imaging algorithms for registration and segmentation tends to develop specialized algorithms for different purposes, especially for image registration. Another approach could be to generalize algorithms but use enhanced knowledge about the underlying image data. To establish this task, a semantically connected information structure is needed to provide algorithms with this data.

The goal of this work is to bring up and to evaluate a generic software system that gathers and stores additional information and builds the ontology as a basis from which further image processing steps can benefit. For this purpose all information will be semantically connected with additional information in a generic manner. Moreover, a simple query interface is provided that allows applying rules on the information to extract semantic connections and knowledge. Furthermore, it is intended to include some image segmentation and image analysis algorithms that initially append information to images added to the database.

Examples for the data stored in this system may be tensors that describe physical properties of the corresponding image; dependencies between images, structures, and between the structures themselves; as well as results of image registration.

Contact: Markus Graf, Ralf O. Floca

Image Registration

© dkfz.de

Image registration is the process of finding a spatial alignment between two images depicting the same or similar scenes. It allows for the fusion of images from different modalities and time points and therefore is an important prerequisite for multimodal diagnostics. Highly specialized algorithms tackle particular registration problems, but they are often only available in specialized research settings; they lack standardized operating environments and independent validation.

The image registration project focuses on several topics that are strongly interconnected. We develop specific registration algorithms, e.g. for lung, prostate, head-neck or paraspinal images involving cancer. A software module for the easy integration of image registration algorithms into a rapid-prototyping platform makes these algorithms available for all applications developed using this platform. Our generic translational registration platform MatchPoint serves as a coordination layer between our algorithms and other applications.

Contact: Ralf O. Floca, Clemens Hentschke, Andreas Jäger, Dörte Corr

RTToolbox

For information about RTToolbox, see this page

Contact: Lanlan Zhang, Ralf O. Floca

AVID

In fractionated radiotherapy monitoring the treatment course of a patient is essential to produce the best possible therapy. Setup errors caused through positioning errors or random errors caused by organ motion are two examples which can lead to deviation errors between the prescribed dose and the actual applied dose. Furthermore, geometrical uncertainties of the tumor make larger safety margins necessary. Monitoring the patients geometrical and dosimetrical variances aid physicians to define compensation strategies.

This work will present a software framework, which allows monitoring geometrical and dosimetrical variances. To analyze geometrical changes of a specific patient or body structure, image registration methods are necessary to address the fact that evaluation of an image registration method has to be done for a specific domain. Thus, an evaluation strategy is presented to quantify the quality of a specific image registration method and to ensure consistent and valid results. To assist specialist in the decision process, graphical result representations are automatically generated for specific time points of a single therapy or for a study. To assist in retrospective analysis in clinical studies, the framework presents graphical data illustration for complete patient collections.

Contact: Andreas Jäger

to top