Medical image processing plays a central role in medical diagnostics and therapy. Multimodal data from a wide variety of imaging procedures, large data volumes due to increasingly higher image resolution and three-dimensional images are just some of the challenges for doctors in differentiated diagnostics. Here, modern image processing provides the key to more efficient and accurate, sometimes even automated, diagnosis and therapy planning.

Technology

At IPE we are engaged in the development of computer-aided diagnostic systems and therapy planning. For this purpose, we do research especially in the field of model-based image registration and automated therapy planning.

Multimodal registration

The comparison and combination of complementary information from different imaging modalities is of great importance both for precise breast cancer diagnosis and therapy planning and for the evaluation of new imaging modalities such as 3D ultrasound computed tomography. However, breast imaging procedures differ considerably in patient positioning and imaging characteristics, e.g. two-dimensional versus three-dimensional imaging, patients in a lying or standing position, imaging of the breast in air or in a water bath.

In order to minimize these differences and thus facilitate multimodal diagnosis, we are researching model-based image registration algorithms that make it possible to predict the spatial relationship, for example, between 2D mammograms and 3D MRI images, 3D ultrasound computed tomography and 3D MRI or 3D MRI and 3D tomosynthesis. The registration is based on a biomechanical model of the breast, which allows finite element (FE) simulations of breast deformations. In addition, we are conducting research on related topics such as image segmentation, optimization methods for model parameters, integration into clinical procedures and computer-assisted detection and classification of tumors.

3D data (here: MRI volumes) are used to automatically generate a patient-specific finite element model of the breast. This is used to simulate a deformation of the breast as it is performed during mammography. The resulting shape of the deformed breast can be directly compared with the corresponding mammogram.

Intuitive visualization of registration results is an essential step to provide radiologists with the greatest possible added value for multimodal diagnosis. We are therefore researching adequate visualization techniques (e.g. image fusion, Figure 2 left) and interactive diagnostic software for multimodal diagnosis (e.g. linked cursor display, Figure 2 right).

Therapy planning

Computer-assisted pre-planning is increasingly becoming a component of patient-specific therapy. Our research in the field of therapy planning mainly covers two fields: planning of breast biopsies and simulation and planning of the application of focused ultrasound.

When planning breast biopsies, we use the development of multimodal registration (see above) to enable therapy guided by a complementary, better available modality even for tumors that are only visible in an imaging procedure. By using model-based image registration, we can, for example, localize a tumor that is only visible in MRI in the X-ray mammogram as well as in the biopsy target image.

For therapy applications with focused ultrasound, we develop simulations of the ultrasound field in order to predict suitable patient-specific application parameters. For this purpose we consider the interactions of ultrasound with soft tissue and bone. To take into account the individual anatomy, we develop (semi-)automated tissue segmentation methods that localize bone and brain in the MRI images.

For students

Students are an essential part of our research group. In our research areas there are continuously exciting tasks for bachelor and master theses, internships and hiwi jobs in various disciplines. Please feel free to contact us directly if you are interested!