Skip to content
27 Jun

Using machine learning to quantify tumour microstructure non-invasively with diffusion-weighted MRI

Named Projects

Biomarkers of tumour grade, prognosis and treatment response in cancer are critically required in the clinic. This project will develop new, non-invasive histology techniques, based on diffusion-weighted magnetic resonance imaging (DW-MRI), that can quantify the microstructure of solid tumors, using machine learning.

This is a 4-year PhD studentship, based at UCL Centre for Advanced Biomedical Imaging (CABI). The funding covers an annual tax free stipend and tuition fees. As the studentship is partially funded by the EPSRC the standard EPSRC eligibility criteria apply, please refer to the EPSRC website for further details. The successful candidate will join the UCL CDT in Medical Imaging cohort and benefit from the activities and events organised by the centre.

DW-MRI uses the diffusive motion of water to non-invasively probe the microstructure of biological tissue. By sensitising the MRI scanner to this diffusive motion, changes in the acquired signal can be mathematically modelled, allowing microstructure to be quantified. Using this approach, we have recently been able to quantify the size and density of tumor cells and blood vessels (1,2).

Our implementation of this ‘non-invasive histology’ technique currently uses simple, geometric models to represent the tumour microenvironment (for example, cells are represented as spheres), in order to make fitting to the data computationally tractable and to remain within the support of the data.

To improve on this, we have been developing a new framework that uses machine learning to define a relationship between DW-MRI signals and the tumour microstructure measurements that they derived from. This approach uses new three-dimensional histology techniques that are being actively developed by our group, and enable us to map the microstructure of complete tumours at sub-cellular resolution. We then use these data to generate synthetic MRI signals, using well-defined physical principals (the Bloch equations). Using machine learning, these data can then be used as complex models to quantify new, in vivo DW-MRI data.

Candidate
Applicants are expected to have an undergraduate degree in Physics, Mathematics, Computer Science or Biomedical Engineering or relevant Physical Sciences based subject passed at 2:1 level (UK system or equivalent) or above. Good working knowledge of C++ and/or Python and/or MATLAB is desirable. The candidate would be provided with extensive training in machine learning, biomedical imaging and image post-processing.

Environment
The project will be undertaken at CABI, which houses a wide range of in vivo and ex vivo imaging equipment to support the work. The candidate would benefit from the unique interdisciplinary environment at CABI, enabling other PhD students and staff to provide peer supervision and assistance with training in cutting-edge techniques.

Application
To make an application for this project please send a CV and cover letter detailing why you want to apply for this studentship, and why you believe you are suited to it, to Dr Simon Walker-Samuel (simon.walkersamuel@ucl.ac.uk).

Application deadline: 11 July 2018. Project start date October 2018.