Optimisation of a pseudo-3D model observer for emerging 2D mammography applications
Model observers are a group of mathematical techniques that are used to predict the detectability of signals in particular background. Present project focuses on signals that have to be detected in x-ray images of the breast. Two types of lesions are important: tiny microcalcifications and cancers.
2D mammography is used in breast cancer screening. Breast tomosynthesis is a new candidate technique that produces a pseudo 3D x-ray image of the breast: when compared to 2D mammography, tomosynthesis reduces the overlap of tissues, but the information is not yet isotropic in 3 dimensions, therefore ‘pseudo-3D’. Recently, 2D images are calculated from tomosynthesis data sets.
In our team, we have successfully developed a model observer of which the scores correlate nicely with the radiologists detecting lesions in breast tomosynthesis data. In present project, the student will tune these algorithms for use in 2D mammography and in synthesized 2D mammograms and will verify whether correlation with human reading is retained. This project is run in a cooperation with the Optimam research group, of which we are partner of.
The project is available in the Fall and Spring semester.
The project is also available for graduate students.
Number of places available: 2 per semester.
- Basic programming skills.
The group is a part of the department ‘imaging and pathology’ of the faculty of medicine, but focuses on medical physics aspects. Most of our research projects start from clinical remarks and/or complaints of the radiologists in the university hospital.
Topics of research include: optimisation of breast cancer screening techniques, patient dosimetry, automated dose and quality monitoring, phase contrast imaging and evaluation of performance, next to MRI research (diffusion weighted imaging, cardiac MRI, fMRI).
Ten Phd students and a similar number of master thesis students work under the guidance of Prof Hilde Bosmans and Prof Nicholas Marshall. Opportunities include the close cooperation with the radiology department, access to emerging x-ray equipment and patient images, a powerful software environment for online patient dose monitoring, a network with 102 mammography units sending data on daily basis for (automated) quality follow-up, a well-developed Monte Carlo framework and experience with virtual clinical trials.