INSPIRE: Integration of Non-linear Sliding Processes into Image REgistration

The purpose of this research project proposal is to design and investigate a new generation of nonlinear image registration methodologies using a novel constrained optimization approach based on the integration of sliding motion of organs. Though completely generic, the approach will be explored primarily for respiratory lung motion compensation in cancer CT imaging, where the sliding motion of the lungs interfacing to the pleura poses a particularly challenging problem in detecting disease and in monitoring of treatment or intervention. We believe that this kind of approach will find application in a wide range of clinical problems where respiratory motion is affecting imaged organs. For example, the slipping motion of masses in breast ultrasound could be tracked for image-guided biopsies. Similarly, liver tumours could be tracked for ultrasound-guided targeted drug delivery, compensating for the sliding of the liver due to the diaphragm contraction during the breathing cycle. One key and entirely novel idea to be developed in this project is to constrain registration to directional sliding motion along the lung surface, while compensating for deformations occurring within the lungs due to expansion or compression occurring at varying levels of respiration. Rather than regularizing the registration cost function locally, a more principled approach will be taken which embeds surfaces of the lungs, and also of nearby organs like the liver, to help drive the registration process in a physiologically plausible manner. Integrating surfaces in the registration will enable us to model - and hence recover - slipping motion of surfaces within an otherwise smooth motion field. Our hypothesis is that this integrated, constrained registration framework will be physiologically more grounded than current, state-of-the-art motion correction approaches which are largely ad-hoc. Consequently the proposed research will provide a significant step towards improved diagnosis and disease monitoring. We will demonstrate the benefit of our new registration methodology by applying it to respiratory motion correction in serial CT lung cancer imaging for patients with malignant pleural mesothelioma, who have been imaged using CT over the course of chemotherapy treatment.

The direct beneficiaries of this research will be lung cancer patients. These benefits include many different aspects: Firstly, the research will contribute toward a significantly improved understanding of lung cancer and treatment response. Monitoring the disease, once detected, is essential for detecting and quantifying changes due to cancer growth or regression, as a response to therapy. The proposed image registration method will be a prerequisite for disease monitoring, and developing and evaluating therapeutic strategies. Secondly, the research will improve patient care by making imaging more robust with respect to patient motion and inconsistent breath holds, and delivering higher quality image data for analysis. Third, since the biophysically-grounded motion discontinuity extension to image registration is quite generic, the research should have impact widely across medical image analysis. To maximise the impact of this project we plan to work closely with our clinical collaborators at the Churchill Hospital, in particular Dr Fergus Gleeson, and will cross-link this research with ongoing research in the Oxford Cancer Imaging Centre, and the Wellcome Centre of Excellence in Medical Engineering at Oxford. This provides an excellent route for translation of the proposed research outcomes into the clinic. We will engage the wider clinical community and the NHS about the impact of these technologies in improving health care provision and reducing costs in the NHS and how they can be widely integrated into clinical workflow. We will submit articles to international journals such as IEEE Transactions on Medical Imaging, or Medical Image Analysis, and to international conferences such as MICCAI and IPMI, or national ones such as BMVA and MIUA. In addition to academic dissemination there will be a prospect of clinical dissemination, including staff groups from radiologists and oncologists. In addition, several companies (including Siemens Molecular Imaginf and GlaxSmithKline) have already expressed strong interest in this research. We will approach these and as well as other companies with a view to explore the potential for commercialization of the developed algorithms. For this we will involve Isis Innovation which is Oxford University's technology transfer company. Isis has a long experience in the commercialisation of intellectual property including patenting, licensing, spin-out companies. We will also involve them in developing an appropriate IP protection strategy for the algorithms developed. Finally, many of the concepts which will be explore in this project have applicability beyond the proposed application in lung imaging. For example, the modelling of discontinuous motion and deformation is crucial in many other medical imaging applications, e.g. in orthopaedic applications involving soft tissues and rigid bones. Similarly, the work on constrained registration will be applicable in many other scenarios.