Medical Imaging and Electronics

At Fraunhofer IMTE, a preclinical MPI system was developed and set up, whose selection field corresponds to a field-free line. This makes it possible to obtain sensitive and high-resolution images of a distribution of magnetic nanoparticles. The system is characterised by mechanical rotation. This makes it predestined for a wide range of preclinical applications, which are being evaluated on this system in collaboration with various partners.

MPI is so far a preclinical imaging modality. At Fraunhofer IMTE, an MPI device is to be developed on the basis of a prototype head scanner, which will be approved for human application and enable the first MPI images of a human being. The envisaged medical application is rapid stroke diagnosis, which can be performed directly at a patient's bedside, e.g. in an ambulance or in the emergency room. This is achieved by designing the system as a mobile unit.

Minimally invasive plate osteosynthesis (MIPO) is the method of choice for bone fracture treatment. By opening only small access points to the surrounding tissue, the osteosynthesis plate and the required screws are brought close to the fracture. MIPO has been shown to shorten recovery times, accelerate fracture healing and reduce the number of infections. The widespread use of MIPO is impeded by the need for radiation-based control for an accurate location, which increases the applied dose. In addition, the soft tissue enclosed in the fracture ends is not visible, which makes MIPO inapplicable and potentially damages nerves and vessels. When placing the screws, no information about the bone quality is available because the bone is not directly visible. In order to counteract the disadvantages, this project aims to build a X-Ray device that significantly increases soft-tissue contrast in 3D and at the same time reduces the patient's exposure to radiation. For these tasks, highly adapted software algorithms are developed to reconstruct the data in a reasonable amount of time. The measurement will allow a direct assessment of bone quality so that the placement of screws can be directed to areas with increased bone density. In order to increase the acceptance of the device, the procedure will be seamlessly integrated into existing treatment processes and workflows of hospitals and surgical centers.

Multi-task Deep Learning for Large-scale Multimodal Biomedical Image Analysis; BMBF-funded

One of the biggest challenges in the field of Computed Tomography is the presence of metal objects such as  hip implants, artificial joints, pacemaker, or orthopedic screws within a patient's body. Caused by the extremely high attenuation coefficients and the physical characteristics of x-ray photons one can obtain severe streaking artefacts in the reconstructed image. The obtained artefacts reduce the diagnostic value of the CT image up to a point where the acquisition is useless for diagnosis. The problem of metal artefact reduction (MAR) has been intensively studied for over 3 decades. A common approach for MAR is to discard the projection data influenced by metal. Therefore, an algorithm must be found that can cope with the gap within the acquired raw data (a sinogram). The majority of the high quantity of different approaches that have been published can be separated in two categories: sinogram completion-based methods and iterative methods.

Our research group is particularly working on algorithms that incorporate criteria for consistent sinograms. An inpainting method that fills the gap in the raw data in a consistent way, regarding these criteria, allows for a reconstruction  that uses all required projections without using the corrupted projections through metal. Another project is focused on the inclusion of prior knowledge in terms of shape and composition of implants into a reconstruction algorithm. The main idea is to use the known attenuation coefficients of the metal implant in order to reduce the streaking artefacts, which are initially caused by the projections that are influenced by metal.