One funded PhD
studentship is available within the University of Kent's Centre for BioMedical
Informatics [1]. The studentship will cover tuition fees for UK or
EU students, and a maintenance bursary (currently at £13,300 per year). The
award will be for full-time study starting in September 2010, and will continue
for three years subject to satisfactory progress. In addition to working on a
chosen PhD research project, the successful student will be required to spend 6
hours per week during the University term time teaching and demonstrating.
Applications from non-EU students are invited, but the student would have to
fund the difference between EU fees and overseas fees themselves if their
application was successful.
A list of projects is given below. All of these projects reflect the
interdisciplinary aims of the Centre, which promotes links and joint projects
between the various disciplines in the Faculty. These projects are based at the
Canterbury Campus,
Qualifications
Applicants
should have a 1st, 2:1 or a Masters degree (or overseas equivalent) in a
subject of relevance to the project being applied for.
How
to apply
Restrictions
Terms and
Conditions
Applicants will
automatically be considered for other studentships offered by the University in
that project area. We would be interested to hear from self-funding students in
these areas - please email [log in to unmask]
or [log in to unmask].
Please
contact the named supervisors with specific queries about the projects
·
Analysis and processing of
Optical Coherence Tomography images for the recognition of eyelid cancer
Supervisors: Prof. Adrian Podoleanu (School of
Physical Sciences; [log in to unmask]
),
Dr. Ali Hojjatoleslami (Biosciences; [log in to unmask] )
Optical Coherence
Tomography is a new high resolution optical imaging that has become a popular
diagnostic tool in ophthalmology, and skin cancer detection. The clinical
application of OCT is rapidly evolving as recent studies demonstrate that this
versatile imaging technology can be used in the clinic to improve the diagnosis
of life-threatening diseases in their early stage [1-5]. It is shown that OCT
images of lesions can be used to detect specific cancer tissue from surrounding
normal tissue. One of the life threatening disease is eyelid cancer which
non-invasive OCT imaging can help the specialists to diagnose the patient in its
early stage of development. Our initial study of eyelid cancer images
illustrate that the pathological signs can be traced on OCT images suggesting
that image processing techniques can be used to detect the cancerous tissue.
This project focuses on the design of an automatic algorithm for the detection
of eyelid cancer from OCT images by applying image computing techniques for
image enhancement, the detection of suspected tissue, and the categorization of
the tissue to cancerous and non-cancerous tissue. OCT images suffer from poor
contrast and high level of speckle noise.
A set of OCT images of eyelid cancer with pathological images and clinical
data, which is obtained by Adrian Podoleanu's group, will be used for this
project. The programme involves the following steps:
This
interdisciplinary project is suitable for a graduate with image processing, or
a closely related discipline with strong programming skills and interest in the
applications of computing in biomedical science. Experience on medical image
analysis applications is an advantage. The student will join an
interdisciplinary team and be supervised jointly by the Neurosciences and
Medical Image Computing group of Biosciences and Applied Optics Group in the
School of Physical Sciences. There will also be collaboration with Dermatology
and Pathology departments of Maidstone and Tunbridge Wells NHS Hospital Trust.
M. Avanaki, S. Hojjatoleslami and A. Podoleanu, "Investigation of
computer- based skin cancer detection using optical coherence tomography,"
Journal of Modern Optics, vol. 56, pp. 1536-1544, 2009.
C. Ahlers, U. Schmidt-Erfurth, Three-dimensional high resolution OCT imaging of
macular pathology, Optics Express, Vol. 17, Iss. 5, pp. 4037- 4045, March 2009.
Julia Welzel, Optical coherence tomography in dermatology: a review, Skin
Research and Technology, Volume 7 Issue 1, Pages 1-9, Jul 2008.
Lingley-Papadopoulos, Colleen A., Image analysis of Optical Coherence
Tomography images of the urinary bladder for the recognition of bladder cancer,
Dissertation, Biomedical Engineering, The George Washington University, June
2009.
Vrushali Raj Korde , Optical Imaging Modalities: From Design to Diagnosis of
Skin Cancer, PhD thesis, University of Arizona, January 2009.
B. Rakesh Penmetsa, OCT imaging for eyelid cancer detection, MSc Thesis, School
of Physical Sciences, University of Kent, UK, 2008.
·
Methods for the detection of Drug
Effects using EEG Data
Supervisors: Prof. Phil Brown (School of
Mathematics, Statistics and Actuarial Science;[log in to unmask] )
and Dr. Ali Hojjatoleslami (Neuroscience
and Medical Image Computing Group; [log in to unmask] )
Summary. Many potential drug targets are localised within the central
nervous system (CNS) and it is therefore critical to understand if, and to what
degree, the drug crosses the blood/brain barrier. However the key problem with
central drug targets is that we cannot directly monitor drug levels in the
brain, so indirect methods are required. One method is to measure the general
activity of the brain via scalp electrodes (the so called electroencephalogram
(EEG)). The EEG is normally used to study proconvulsant activity or to study
sleep changes. "Snap shot" signatures in rats have been noted by
Dimpfel (2003), but these are not time-varying, but other work has shown that
the EEG signal contains a "signature" which can be used to directly
measure the activity of drug in the brain over time.
Pfizer's current studies use a telemetrised Rat model, which yield a
single-channel EEG signal, and data of this form will be available for the
student to develop analysis and detection methods. They may also be able to
access to EEG data from human subjects, which is more complex, with signal from
multiple channels. If they have access to this data the student will be able to
extend the methods used for the simpler rat EEG signals to the more complex
human case. (See Saletu et al (2002) for previous work in this area).
The student will be working in a novel area needing primary statistical/signal
analysis skills, and will gain experience in the understanding of
electrophysiological measurements, and analysing EEG signal. The analysis
skills developed and learned in the analysis of EEG data will be applicable to
other high dimensional data problems. The analysis of high dimensional data is
increasingly important, and poses a wide range of interesting questions.
Non-invasive biological markers of brain activity such as EEG data are
important in the drug discovery process, and the student should be able to make
a valuable contribution to improving these processes. The methods being
developed are also translatable, as they can be used on both animal and man.
Aim EEG signals will change depending on the status of the animal/human,
as brain-wave patterns change markedly during deep sleep, sleep, active state,
and resting. In rats, in particular, transition between the different states
can occur many times during a typical 12 hour study. For a given drug type we
are looking for a "signature"(a profile across frequencies) in the
EEG signal which is independent of the sleep state. The amount of this profile
present in a given spectrum at a given time will reflect the affect of the drug
present in the brain. We expect these signals to follow a PK/PD model, but may
not necessarily know the exact parameters, so these will need to be estimated
as part of the signature detection.
Detecting this signature could provide an important biomarker for the drug
discovery and development process. It will provide a non-invasive way of
detecting drug effects that is translatable, so can be used in both animal and
human studies.
Possible Analysis Methods. The data format takes the form of a matrix or
a three dimensional array, so multivariate methods are expected to be useful (a
relevant approach to the three dimensional case is Linder and Sundberg, 2002,
and references therein). Pfizer's current approach uses a modified version of
Canonical Correlation Analysis (MDCCA). A novel part of the study is that we
have a functional form for the response we are trying to detect, so methods
which are capable of detecting this underlying functional form and quantifying
it (eg Ramsay and Silverman, 1997) will also be relevant. The MDCCA simply
optimises the correlation across time between a linear combination of the power
spectrum and the standard PK/PD curve parameterised according dose level. There
are questions of robustness, alternatives to the simple two compartment PK/PD
model, diagnostics for detecting inadequacies in the model and overall
inferential properties both classical and Bayesian.