Contouring & Segmentation

Segmented structures play a crucial role in the RT field. Contoured structures delineating target volumes and organs at risk are needed for the planning process, as they allow for precise evaluation of the dose distribution using dose-volume histograms. Manual segmentation of all relevant structures is a tedious and time-consuming task. Thus, there is a need for fast and robust tools supporting physicians in segmentation tasks.

Fraunhofer MEVIS has developed a number of modules enabling creation and editing of segmented structures. The user can choose the representation used to store segmented structures, either a label image, planar contours or a mesh.

One of the tools developed within the SPARTA project called Contouring with Snapping and Interpolation (CSI) aims not only at speeding up the manual contour correction but also at quick from scratch segmentations. CSI offers different clinically well-established interactive tools, such as ball tool with adjustable radius, or a freehand tool. The tool operates with a sub-voxel precision and offers a snapping mode for contour refinement using local image information. This reduces the required degree of precision, as the resulting contours are automatically aligned with the underlaying image.

Additionally, CSI implements a sophisticated workflow involving automatic contour interpolation and extrapolation to support fast generation of new structures. The tool keeps track of which contours have been edited manually and attempts to inter- or extrapolate additional contours to fill up empty space. Once an automatic contour is corrected by the user, it becomes a manual one and will no longer be modified by the automatic interpolation algorithm.

Extensible Software Framework for adaptive Radiotherapy

Fraunhofer MEVIS is developing an application framework allowing for rapid prototyping of clinical applications. The basis of this technology is MeVisLab, a rapid prototyping platform for medical image analysis. Our goal in SPARTA is to provide a solid application infrastructure combining the image-processing and visualization capabilities of MeVisLab, with modern web-based application design.

Non-linear Image Registration

Fraunhofer MEVIS is developing problem specific, highly efficient methods for image registration in the context of interfractional adaption of radiotherapeutic treatment for paraspinal or head- and neck tumors. These methods are optimized for the adaption of interfractional varations. They are capable of detecting and correcting for positioning inaccuracies and can hint at requirements for plan-adaptations. Additionally, they build the basis for non-linear transfer of segmented structures between fractions to support re-planning during the course of a therapy.

To this end, a fully automated non-linear registration workflow for head and neck CT data has been developed. It was integrated into a software prototype for automated contour transfer, which is currently evaluated at the hospital of the Ludwig-Maximilians-University (LMU) Munich. Additionally, an adapted solution focusing on the registration of cone-beam CT images in under development, exploring the possibilities of using CBCTs for dose-accumulation, by means of non-linear mapping to the planning CT.

User Experience (UX)

User Experience Engineering (UX) revolves around the many aspects of how humans operate and interact with computer systems. Specifically, it puts the focus on how the user feels when doing so, how intuitively and how efficiently a given goal can be reached. In SPARTA, UX goals aimed at providing usability concepts for typical workflows involved in daily routine in radiotherapy. Requirements were analyzed through interviews and job shadowing in daily clinical routine, as well as during hands-on studies at various workshops. Based on the acquired data, a concept for a primary workflow ‚planned vs delivered dose‘ was developed. The concept was iteratively refined evolving from initial paper drafts, over functional wireframes up to a prototypical implementation.


Coming soon.