Imperial College London Featured PhD Programmes
Imperial College London Featured PhD Programmes
The Francis Crick Institute Featured PhD Programmes
Norwich Research Park Featured PhD Programmes
Cardiff University Featured PhD Programmes

Convolutional neural networks to correct motion in dynamic MR imaging

This project is no longer listed on FindAPhD.com and may not be available.

Click here to search FindAPhD.com for PhD studentship opportunities
  • Full or part time
    Dr L Kershaw
    Dr S Semple
  • Application Deadline
    No more applications being accepted
  • Funded PhD Project (European/UK Students Only)
    Funded PhD Project (European/UK Students Only)

Project Description

This project is one of 9 four-year PhD Studentships funded by Medical Research Scotland (http://www.medicalresearchscotland.org.uk) to be delivered jointly by the named University and Company. The Studentship will provide the first-class academic and commercial training needed to equip the successful candidate for a science career in an increasingly competitive market.

The project is delivered by the University of Edinburgh [Supervisors: Dr Lucy Kershaw and Dr Scott Semple (both Edinburgh Imaging)] and Canon Medical Research Europe Ltd (https://research.eu.medical.canon) [Company supervisor: Dr Keith Goatman]. You will spend half your time with the Edinburgh Imaging and half based within the AI Research Team at Canon Medical Reserch Europe.

Magnetic resonance imaging gives excellent views of the structure of soft tissues in the body. It can also measure how well these tissues are functioning. Dynamic contrast enhanced (DCE) MR is one way of achieving this by injecting a small amount of “contrast agent” (so-called because it shows up brightly on images) before acquiring a series of images. The resulting sequence of images show the contrast agent washing in, and then washing out, of tissue. The rate at which the contrast agent enters and leaves tissues provides important information about the health of the tissue, and these parameters may be calculated for every pixel in the image.

Motion can be a severe problem with dynamic scans. For DCE MR to work, each pixel must align with the same bit of tissue in every image in the sequence. The analysis will fail if a pixel in the first image is liver tissue but then the patient moves and that same pixel is now located in the lung in the next image.

There is a class of algorithms, known as image registration, that automatically align image sequences. However, dynamic contrast-enhanced image sequences are particularly challenging as the image appearance changes due to the contrast uptake, so a dark pixel in one image could be a bright pixel in the next, or of course that difference could just be due to movement. Additional complexity comes from the non-rigidity of the abdomen and various artefacts that corrupt MR images. To achieve state-of-the-art high precision, fast and accurate motion correction the most promising current solutions are based on convolutional neural networks.

Motion can be categorised by its speed relative to the image acquisition. Fast impulse motion (e.g. coughing) will result in image artefacts, while slower motion (e.g. gentle breathing, bladder filling) will result in misalignment between the images in the sequence. The first goal will be to correct for slow motion misalignment between frames, then to correct for image artefacts due to rapid intra-acquisition movement.

Clinical applications of DCE MR include imaging the uterus, liver or prostate gland. For the uterus, a major source of motion is the bladder, which can double in volume during scanning – lifting and rotating the uterus. The uterus is, therefore, a good model case to develop a broader motion correction method. There have been no dynamic uterine studies to date that explicitly corrected for bladder motion.

Canon Medical Research Europe Ltd are based in Edinburgh. They design and develop leading edge software for medical image visualization and analysis used in CT, MRI and other medical scanners used in hospitals worldwide. They are an R&D centre of excellence with strong academic and research links.

ENQUIRIES:

Enquiries should be sent by email to Dr Lucy Kershaw:
[Email Address Removed]

APPLICATIONS:
Candidates must have obtained, or expect to obtain, a first or 2.1 UK honours degree, or equivalent for degrees obtained outside the UK, in a relevant discipline (eg physics, engineering, computer science, mathematics) and demonstrated ability writing software. A Masters level degree in relevant topic is desirable, as is experience with deep learning and convolutional neural networks or image analysis experience. Above all candidates must be self-motivated and intellectually curious.

Applicants should send a CV and a covering letter, describing their research interests and explaining why they wish to carry out this project, by email to Alexandra Moreira (Postgraduate Administrator):
[Email Address Removed]

Applicants must arrange for two academic referees to submit letters of reference via email to Alexandra Moreira:
[Email Address Removed]
before the closing date for applications.

Short-listed candidates will be notified by email.

Your application may be shared with the funders of this PhD Studentship, Medical Research Scotland and Canon Medical Research Europe Ltd.

Interviews are expected to take place approximately 3-4 weeks after the closing date for applications.

The PhD Studentship start date is negotiable.

Funding Notes

PhD Studentship provides:

An annual tax-free stipend of £17,500, increasing to £18,000 over the four years; tuition fees at UK/EU rates only; consumables; and contribution to travel expenses. International fees are not covered.

References

www.ed.ac.uk/edinburgh-imaging

How good is research at University of Edinburgh in Clinical Medicine?

FTE Category A staff submitted: 206.93

Research output data provided by the Research Excellence Framework (REF)

Click here to see the results for all UK universities


FindAPhD. Copyright 2005-2019
All rights reserved.