Engineering Doctorate (EngD) - Registration in Medical Imaging

Industrial Partner: Toshiba Medical Visualization Systems Europe

Image registration concerns discovering a spatial transformation to align corresponding features in 2D or 3D images. It has become a crucial enabling technology in medical imaging, for applications such as aligning images from a current and previous scan of a patient in order to understand disease progression or efficacy of treatment. Inter-modal registration involves the alignment of e.g. an MR and a CT scan from a patient allowing decisions to be based on the fused result. In population studies, or where aligning an 'atlas', the registration problem is inter-patient, posing particular problems due to anatomical variation. Some applications concern alignment with images acquired during an interventional procedure, imposing requirements for "real time" performance. The aligning transformation required can be a simple translation, or fully elastic warping, depending on the application. In some cases, the purpose is alignment for direct presentation to clinicians, though increasingly the result feeds into further image analysis algorithms.

Thus there are many challenges, arising from changing pathology, varying anatomy, artifacts and noise particular to each imaging modality and the requirement for rapid execution on large datasets. This project will tackle these issues with the goal of creating a flexible registration framework for medical applications.

The student will require core skills in mathematics, numerical optimization, image processing and software engineering. He or she will work as an integrated member of the TMVS Research Team.

We are looking for a computer scientist, engineer with signal processing experience or a physicist/mathematician with a strong programming track record. So the ideal candidate would have a combination of: computer science, mathematics & statistics, vision & signal processing, and physics.