Copy of Digital Breast Tomosynthesis

We propose a project that combines the skills of the imaging groups at UCL and Oxford with the expertise of Dexela, a UK SME dedicated to the development and commercialisation of DBT that was co-founded by leading researchers in DBT from the US. The project will be managed by Dexela. Our clinical advisory group will comprise radiologists from King's College Hospital, the Royal Marsden Hospital, and St. Bartholomew's Hospital, all in London, together with The Massuchusetts General Hospital in the US. The overall aim of our project is to establish Digital Breast Tomosynthesis (DBT) as the modality of choice in breast cancer detection by enhancing its sensitivity and specificity sufficiently to create compelling clinical and economic benefits. We will first establish DBT for difficult cases (e.g. the dense breast, previous surgery, younger women, implants and suspicious regions in inaccessible areas), while aiming ultimately to replace mammography in national screening programmes. This project combines innovative work in optical imaging of the breast and neo-natal brain (Arridge), in mammography image processing (Brady), in registration and change detection (Hawkes) and DBT (Dexela). While some of the core components already exist in isolation, this project is the first attempt at such a combination and novel contributions are expected in all the areas. The new DBT image acquisition device developed by Dexela will have a superior geometry to existing systems and significant flexibility in acquisition geometry, breast compression and exposure parameters per view. The project will explore the optimisation of acquisition parameters informed by sophisticated image processing. The innovative steps concern: i) iterative reconstruction methods applied to very large DBT datasets (over 1 gigabyte) verses filtered back projection (FBP), ii) the wide 90 degree angular range (verses 16-50 degrees for the majors) and variable angular spacing of projections iii) varying voltage, current and detector resolution between projections (verses invariant parameters) iv) the low number of projection images (11 verses 15- 48), which results in faster acquisition, less patient movement and better signal to noise ratio, v) the implementation of Intelligent Image Acquisition which provides real-time feedback to the image acquisition parameters, and vi) the innovative combination of 4D reconstruction methods, and change detection. The work on registration and automated comparison in DBT is innovative and there is no similar research which has been reported related to DBT. This innovative work may have application beyond this field in other medical and non-medical applications.