Registration with constraints Radiation Therapy
© Fraunhofer MEVIS, Image courtesy of Inselspital Bern
Registration of CT-scans with disabled non-rigid deformations for bone structures. The degree of non-rigidity of the deformation field is color-coded.

Image Registration

Respecting anatomy: Disabled non-rigid deformations for bone structures. The degree of non-rigidity of the deformation field is color-coded.

Liver Tissue Registration of Serial sections
© Fraunhofer MEVIS
The registration of serial sections is a key technology in digital pathology. The image shows two virtually-combined serial sections, stained with Hematoxylin and Eosin (H&E) and Glutamin-synthetase (GS), respectively. The H&E stain makes numerous fat droplets visible that indicate a pathology of the liver. The GS stain highlights central veins with a brown rim, so that they can be distinguished from colorless portal fields. The virtual combination of both stains reveals that, in this case, steatosis is concentrated around portal fields but not around central veins.

Image Registration

Virtual multi staining: Combination of differently stained tissue sections for more extensive analyses.

Multi-modal Registration brain neck carcinoma
© Fraunhofer MEVIS
Multi-modal registration to examine an head and neck carcinoma with different medical imaging techniques.

Image Registration

All information in one place: Fusing morphological and functional information for better health care. Left to right: Histology and Mass Spectrometry, CT, MR, and PET.

Image Registration


Mono-Modal and Multi-Modal Registration

We develop sound and efficient solutions for rigid, affine, and deformable registration for a wide range of applications and image types. Supported modalities include CT, MR, PET, SPECT, interventional modalities such as CBCT and US, histological H&E and immunohistochemical stains, as well as preclinical and biological modalities like spectral MALDI imaging and light sheet fluorescence microscopy (LSM/SPIM).

Flexible Framework

We developed a unified, flexible, and extensible theoretical framework for variational image registration that is independent of third-party software. Our methods build on a strong mathematical background. This includes various aspects from applied mathematics like mathematical modeling, variational approaches, partial differential equations, scientific computing like numerical optimization, numerical linear algebra, high performance computing, image and signal processing.

Tailored Solutions

We develop and put in practice methods for modeling application conformal constraints into registration. Examples are keeping a-priori known corresponding landmarks, locally rigid motion of bones, volume preservation of tumors, or guaranteed topology preservation.

Efficient Implementations

Speed matters. Besides efficient optimization and numerical schemes we are taking full computational advantage of available standard hardware such as multi-core CPUs and GPUs. We develop and provide efficient parallel implementations purely based on CPUs as well as solutions that run with the power of GPUs.

Gigapixel Image Registration

Computational pathology and biology are rapidly emerging fields that request processing extremely large images. The size of single images ranges from gigabytes up to terabytes. We address these computational challenges and develop techniques that allow fast and highly accurate registration of 2D and 3D gigapixel images.

Flexible Integration

We provide registration software to our partners on all levels ranging from functional research prototypes to efficient quality-assured (EN ISO 13485 and EN 62304) components for use in medical devices. Typical products are C/C++ libraries (DLLs), command line tools, or standalone applications available for all major operating systems. Our software is designed for easy integration in our partners' environments.


© Fraunhofer MEVIS, Image Courtesy of Inselspital Bern

Motion-Correction & Tracking

© Fraunhofer MEVIS

Multi-Modal Image Registration for Large Images in Computational Pathology

© Fraunhofer MEVIS

3D Reconstruction of Multi-Stain Histology