Panel for Biomedical
Engineering
Young researchers futures meeting
Neural engineering
19-21 September 2012
Panel for Biomedical
Engineering
Dear Colleague
On behalf of the Panel for Biomedical Engineering I would like to welcome you to the University of Warwick and the
2012 Royal Academy of Engineering, Young Researchers Futures Meeting on Neural Engineering.
The Panel for Biomedical Engineering (hosted by The Royal Academy of Engineering) provides a forum through
which the principal organisations concerned with biomedical engineering can communicate, debate and work
together to improve the diagnosis and treatment of major medical conditions. The Young Researchers’ meetings
aim to bring together the most excellent early-career researchers to showcase their work, network and receive
guidance from leading experts.
Neural engineering is a field that is now highly popular, both as a field of research as well as in developments of
real-world products being used to enhance diagnosis and prognosis, as well as enhancing people’s lives. In Neural
Engineering, engineering skills and techniques are used to understand, repair, replace, or enhance neural systems.
Neural engineers solve design problems at the interface of living neural tissue and non-living systems. Like other
areas in biomedical engineering, neural engineering draws on the expertise of a variety of disciplines such as
computational and experimental neuroscience, clinical neurology, electrical engineering and signal processing.
Of special note is the interface to living neural tissue, which brings together elements from robotics, cybernetics,
computer engineering, neural tissue engineering, materials science and nanotechnology. Neural engineering still
has a strong research component but has some very clear and beneficial goals. These are centred on the restoration
and augmentation of human function via direct interactions between the nervous system and artificial devices,
for example, restoring the ability to interact with the environment following a stroke or debilitating neural disease
through interfacing technology directly with the human nervous system. It is also possible to augment human
function in much the same way.
The meeting at the Institute of Digital Healthcare, University of Warwick will bring together young researchers
and mentors in varied areas and disciplines underpinning neural engineering, from neural interfacing, to neural
modelling, computational intelligence, neurorehabilitation and neural networks/ cybernetics. We hope that you
use this opportunity to interact with your peers and debate the field enhancing your mutual experience of neural
engineering.
On behalf of the Panel for Biomedical Engineering, we wish you a highly successful and enjoyable conference.
Yours sincerely,
Professor Christopher James
Royal Academy of Engineering Panel for Biomedical Engineering
Programme
DAY ONE - 19 September 2012
1:00pm
Lunch
1:30pm
Start
Boardroom
Welcome & overview: Professor Christopher James
Session one: neural signal processing and modelling
1:45-2:30pm
Expert talk
Towards interpretable classifiers using blind signal separation
Professor Paulo Lisboa, Liverpool John Moores University
2:30pm
Vaibhav Gandhi (University of Ulster)
EEG filtering with quantum neural networks for a Brain-Computer Interface (BCI)
3:00pm
Simon Davies (University of Warwick)
Empirical mode decomposition for feature extraction in P300 responses for BCI
3:30pm
Samantha Simons (University of Surrey)
Fuzzy entropy and multiscale fuzzy entropy of the electroencephalogram in
alzheimer’s disease
4:00-4:25pm
Refreshment break
4:25pm
Simao Laranjeira (University of Oxford)
Modelling glutamate dynamics and associated tissue temperature change in
brain cells after stroke
4:55pm
Arsam Shiraz (University College London)
Design and development of a closed-loop active neuromodulation device
for treating urinary incontinence
5:25pm
Close
(Check in to rooms Rootes)
7:00pm
Dinner
Rootes restaurant
Royal Academy of Engineering 1
DAY TWO - 20 September 2012
Session two: Biological/technical interfaces
8:30am
Tea, coffee
9:00am
Boardroom
9:00-9:45am
Expert talk
Reading and writing signals from and into the brain optically
Dr Simon Schultz, Imperial College London
9:45am
Julian George (University of Oxford)
One network to another - engineering a neural interface through micro-porous membranes
10:15am
Deren Barsakcioglu (Imperial College London)
Ultra-low-power real-time spike sorter for neural prostheses
10:45-11:00am
Refreshment break
11:00am
Lionel Chaudet (Imperial College London)
Improving opto-electronic neural stimulation with micro-LED arrays and optics
11:30am
Sivylla-Eleni Paraskevopoulou (Imperial College London)
Ultra low implantable platform for next generation neural interfaces
12:00pm
Vasiliki Giagka (University College London)
An implantable stimulator system for neuro-rehabilitation in paralyzed rats
12:30-1:15pm
Lunch
Session three: neural prosthesis
1:15-2:00pm
Expert talk
Implantable neural interfaces for prosthesis control
Dr Ed Tarte, University of Birmingham
2:00pm
Ian Williams (Imperial College London)
Proprioceptive feedback for prosthetic arm users
2:30pm
Graeme Coapes (Newcastle University)
Silicon neuron node capable of large-scale simulations of bio-physically realistic ion channels
3:00-3:15pm
Refreshment break
3:15pm
Jun Wen Luo (Newcastle University)
A novel hardware architecture for large scale hybrid bio – silicon network
3:45pm
Richard Barrett (University of Birmingham)
Spiral peripheral nerve interface: a novel interface design
Session four: neural communication
4:15-5:00pm
Expert talk
Signal processing challenges in neural communication (BCI)
Professor Christopher James, University of Warwick
5:00pm
Balazs Szigeti (University of Edinburgh)
The virtual C. elegans and the behavioural Turing test
5:30pm
Close of Session
7:00pm
Dinner
Scarman House
Royal Academy of Engineering 2
DAY THREE - 21 September 2012
Session five: functional neuroimaging
8:30am
Tea, coffee
9:00-9:45am
Expert talk
Modelling and methods to control the field of activation
for deep brain stimulation
Professor Richard Bayford, Middlesex University
9:45am
Andrew Stewart (University of Edinburgh)
Single-trial prediction of visual perception from EEG: analysis, classification
and ICA source dynamics
10:15-10:30am
Refreshment break
Session six: neural informatics
10:30-11:15am
Expert talk
Neuroinformatics
Professor David Willshaw, University of Edinburgh
11:15am
James McGuinness (University of Stirling)
Role and implications of population heterogeneity on vestibular ocular
reflex response fidelity
Session seven: Cognitive neurology
11:45-12:30pm
Expert talk
Auditory cognition and neural engineering
Professor Tim Griffiths, Newcastle University
12:30-12:50pm
Discussion
12:50pm
Session close
1:00pm
Lunch
2:00pm
Depart
Royal Academy of Engineering 3
Delegates list
Mentors
Professor Richard Bayford MIEEE FInstP FInstIPEM
Middlesex University
Professor Tim Griffiths FMedSci
Newcastle University
Dr James Harte
University of Warwick
Professor Christopher James
University of Warwick
Professor Garth R Johnson FREng CEng FIMechE
Newcastle University
Professor Paulo Lisboa
Liverpool John Moores University
Dr Simon Schultz
Imperial College London
Dr Ed Tarte CPhys FInstP
University of Birmingham
Professor David Willshaw FRSE FInstP
University of Edinburgh
Researchers
Richard Barrett
University of Birmingham
Deren Barsakcioglu
Imperial College London
Lionel Chaudet
Imperial College London
Graeme Coapes
Newcastle University
Simon Davies
University of Warwick
Vaibhav Gandhi
University of Ulster
Julian George
University of Oxford
Vasiliki Giagka
University College London
Simao Laranjeira
University of Oxford
Jun Wen Luo
Newcastle University
James McGuinness
University of Stirling
Sivylla-Eleni Paraskevopoulou
Imperial College London
Arsam Shiraz
University College London
Samantha Simons
University of Surrey
Andrew Stewart
University of Edinburgh
Balazs Szigeti
University of Edinburgh
Ian Williams
Imperial College London
Royal Academy of Engineering 4
Mentor biographies
Royal Academy of Engineering 5
Neural engineering - mentor biographies
Professor Richard Bayford MIEEE, FInstP, FInstIPEM
Middlesex University
Head of Biophysics at the Middlesex University Centre for Investigative
Oncology, Professor of Bio-modeling and Informatics and Head of Departmental
Research, Middlesex University and Honorary Senior Lecturer, Department of
Electrical and Electronic Engineering, University College London.
Bayford’s expertise is in biomedical imaging, bio-modelling, electrical
impedance tomography (EIT), nanotechnology, deep brain stimulation, telemedical systems, instrumentation and biosensors.
He pioneered the first reconstruction algorithm to image impedance changes
inside the human head. He has worked as PI on many EPSRC, EU and industrial
sponsored research projects.
He has been the director of a major EPSRC Nanotechnology Grand Challenges
Healthcare project (EP/G061572; “New imaging methods for the detection
of cancer biomarkers” £1.7M). This project involved collaboration with
Midatech Ltd. (a leading nanotechnology company headed by Professor Tom
Rademacher) and Zilico Ltd.; Midatech are now funding one of our research
projects and Zilico are providing equipment. He co-ordinated a major tentel EU project, Medical Diagnosis, Communication and Analysis Throughout
Europe (MEDICATE) to develop a system to identify links between Asthma
and air quality. This project included the two companies, Jarger (now known
as Carefusion) and Cable and Wireless. He has also had past collaborations
with Medtronic who have provided DBS electrodes for research projects. His
principle area of research focuses on the development of image reconstruction
algorithms and hardware development for imaging brain function. He has
had long collaborations with multidisciplinary research groups both in the UK
and overseas on biomedical applications of EIT and bio-impedance. He has
published over 200 scientific papers.
He has been guest editor on three special issues and co-organizer of three
conferences on biomedical applications of EIT. He is the Editor-in-Chief of the
IoP Physiological Measurements journal, a member of the editorial board of
the International Journal of Biomedical Imaging and Chairs the Publication
committee for IPEM.
Title of presentation
Modelling and methods to control the field of activation
for deep brain stimulation
Abstract
Despite the widely accepted clinical efficacy of deep brain stimulation (DBS),
the underlying physiological mechanisms of this therapeutic tool have not
yet been fully discovered. This has greatly limited the development of more
efficient and safer DBS systems, which, in turn, has not provided researchers
with better means to explore the effects of DBS. This presentation describes
the development of methods to model the electrical field in the human
head and achieve better selectivity and electric field shifting during deep
brain stimulation. Methods based on the use of a current-steering tripolar
electrode configuration, characterized by tunable ratio, 움, between the anodic
currents, combined with the use of triangular current pulses are explored and
illustrated. The result is to minimize the anodal break excitation associated
with fast decays of square pulses.
Finite element models were employed in order to investigate the behavior of
the electric fields generated with the tripolar electrode configuration. Physical
models were also developed to validate the electric field profiles and shifting
capability of the adopted tripolar configuration.
Royal Academy of Engineering 6
Neural engineering - mentor biographies
Professor Tim Griffiths FMedSci
Newcastle University
Timothy Griffiths is Wellcome Senior Clinical Fellow and Professor of
Cognitive Neurology at Newcastle University. His research concerns human
complex sound processing; the analysis of auditory patterns relevant to
speech, music and environmental-sound. He studies deficits in complex
sound processing in patients with brain lesions, functional imaging data
(fMRI and MEG) from normal subjects, and depth electrode data from the
auditory cortex of neurosurgical patients. The functional imaging is carried
out at the Wellcome Trust Centre for NeuroImaging in London, where he
is a Principal, and the depth electrode data is acquired at Iowa where he is
adjunct Professor. These studies allow inference about normal complexsound processing mechanisms. Other work explores abnormal complexsound analysis in developmental and degenerative disorders, and brain
mechanisms for tinnitus and auditory hallucinations.
Title of presentation
Auditory cognition and neural engineering
Abstract
My group works on auditory cognition: the process by which the brain makes
sense of the complex patterns in the acoustic world and the way in which
this can go wrong in brain disorders. I will describe how the brain ‘abstracts’
perceptual cues like pitch and analyses auditory objects based on functional
imaging data and the study of patients with brain disorder.
Auditory cognition is an area of great potential application of neural
engineering. The processors for digital hearing aids and cochlear implants
are primarily targeted to improve speech perception whilst pitch perception
and the recognition of non-speech objects have not been such an emphasis.
Tinnitus can be considered a disorder of auditory cognition for which
brain interventions like transcranial magnetic stimulation, direct current
stimulation and even surgery are currently being evaluated.
Royal Academy of Engineering 7
Neural engineering - mentor biographies
Professor Christopher James
University of Warwick
Professor James is a biomedical engineer and his research activity centers on
the development of biomedical signal and pattern processing techniques, as
well as the use of technological innovations, for use in advancing healthcare
and promoting wellbeing. Neural Engineering forms a large part of his work,
as to date his work has concentrated on the development of advanced
processing techniques applied to the analysis of the electromagnetic
activity of the human brain, primarily in Brain-Computer Interfacing. Prof
James has published over 160 papers in neural engineering in varied
biomedical engineering journals and refereed conferences.
He is immediate past-chairman of the Institute of Electrical and Electronic
Engineers (IEEE) UK & Republic of Ireland (UKRI) Section, Chair of the
IEEE UKRI Engineering in Medicine & Biology Society (EMBS) Chapter,
and a member of IEEE the EMBS Administrative Committee as Europe
Representative. He is on the council of the European Alliance of Medical
and Biological Engineering Societies (EAMBES). He is past Chairman of the
Executive Committee of the Institution of Engineering and Technology (IET)
Healthcare Technology Network, is on the IET TPN Steering Committee and
has advised IET on Healthcare Technology matters for the Faraday Lectures,
and has presented for IET at outreach activities. He is also on the Royal
Academy of Engineering’s Panel for Biomedical Engineering.
Professor James was founding Series Editor for the Biomedical Signals and
Systems book series of Artech House Publishers; Editor in Chief of the Open
Medical Informatics Journal, and is Associate Editor for IEEE Transactions on
Biomedical Engineering and sits on the editorial advisory board of the IEEE
Spectrum Magazine. He is Associate Editor of the IEEE EMBS Conference
Editorial Board (Neural Engineering Theme).
Title of presentation
Signal processing challenges in neural communication (BCI)
Abstract
It is quite clear to everyone in the BCI community that EEG-based BCI
systems require relatively simple and inexpensive equipment, should be
portable and relatively easy to set up, and thus should function in most
real-world environments. Some specific features (such as P300s, visually
evoked potentials (VEPs), cortical neuron activity, μ rhythms, and others)
can then be extracted from the raw EEG recordings in order to form a
meaningful interface. However, this is a non-trivial problem and presents
several challenges. The presentation is centred on the challenges presented
in designing and delivering such a functional real-world system. In general,
the aim is to realize a unique, non-invasive and simple to use system that
is able to offer “locked-in” patients (and others) the partial restoration of
communication and control capabilities to significantly enhance their quality
of life. The main thrust of this research is to take BCI out of the lab and
into the home. This entails building a practical system that is beneficial to
the users and easily manageable by their care-givers. For such a practical
system we must target all aspects of BCI research in order to establish a
patient-driven BCI system.
Royal Academy of Engineering 8
Neural engineering - mentor biographies
Professor Paulo Lisboa
Liverpool John Moores University
Paulo Lisboa studied Mathematical Physics at Liverpool University, taking
a PhD in theoretical particle physics in 1983. He is Professor in Industrial
Mathematics and heads the Department of Mathematics and Statistics
in the School of Computing and Mathematical Sciences at Liverpool John
Moores University.
He holds cross-Faculty positions as chair of the executive committee of the
Centre for Health and Social Care Informatics and co-lead of the Medicine
and Therapeutics Network in the Institute for Health Research.
His research is focused on computer-based decision support in healthcare,
the analysis of public health data for policy reporting and commissioning
purposes, and computational data analysis in a range of applications
including sports medicine and computational marketing. Particular aspects
include principled approaches to case-based retrieval of reference cases,
source identification in Magnetic Resonance Spectroscopy, flexible models
for hazard estimation following surgery for breast cancer and scalable
conditional independence maps for multimodal data fusion.
He has over 200 refereed publications with awards from the journal Neural
Networks for most cited article 2006-10 and most downloaded article in
2003.
He chairs the Medical Data Analysis Task Force in the Data Mining Technical
Committee of the IEEE-CIS and is Associate Editor for Neural Networks, IET
Science Measurement and Technology, Neural Computing Applications,
Applied Soft Computing and Source Code for Biology and Medicine. He is
also a member of the EPSRC Peer Review College and an expert evaluator
for the European Community DG-INFSO.
Title of presentation
Towards interpretable classifiers using blind signal separation
Abstract
Blind signal separation is a powerful tool to represent complex signals
as linear combinations of component sources. This talk will review a
methodology to assign class membership directly from the elements of
the mixing matrix, using a semi-supervised approach. Starting with a
probabilistic classifier, a localised Fisher information measure is derived
which defines a natural metric in data space. This Riemannian metric is then
embedded in a Euclidean space so that convexNMF can be applied. Results
with synthetic data models of multi-voxel Magnetic Resonance Spectra
show that the classification accuracy of the method is competitive and the
composition of individual voxels can be interpreted as a mixture of labelled
component sources.
Royal Academy of Engineering 9
Neural engineering - mentor biographies
Dr Simon Schultz
Imperial College London
Simon Schultz is Senior Lecturer, and Royal Society Industry Research
Fellow, in the Department of Bioengineering at Imperial College London. He
trained in physics and electrical engineering, before completing a DPhil in
computational neuroscience at Oxford University in 1998. This was followed
by postdoctoral stints in experimental neuroscience with Tony Movshon at
New York University, and Michael Häusser at UCL. He joined Imperial College
in 2004, and has led the development of Imperial’s critical mass in the area
of Neurotechnology.
He is widely known for work on neural coding. He has been amongst the
pioneers in the use of two-photon imaging to study neural coding and
has also worked on large-scale computational models of cortical circuits.
He is the Chair of Imperial’s Neuroscience Technology Network, and PI of
the Imperial College node of the new EU Marie-Curie Training Network
“Neural Engineering Transformative Technologies”. He is Associate Editor
for the Journal of Computational Neuroscience, and serves on the National
Committee of the British Neuroscience Association as Membership
Secretary.
Title of presentation
Reading and writing signals from and into the brain optically
Abstract
Recent developments in multiphoton microscopy techniques have
revolutionized our ability to record, and more recently perturb, functional
signals from identified elements of the mammalian cortical circuit. In this
talk I will discuss recent work on using multiphoton calcium imaging to
image functional signals at subcellular resolution in vivo [1]. To date our
work has largely made use of AM-ester calcium dyes such as Oregon Green
BAPTA-1 AM, which still shows the largest signal to noise ratio for detection
of action-potential induced calcium influx. However, the advent of Crelox technology combined with improved genetically encodable indicator
proteins now enables us to target reporters of functional neural signals
to individual cell classes. Another development of interest is the use of
optogenetic modulation to perturb individual elements of the cortical circuit
– I will discuss our recent work making use of the light-activated proton
pump ArchT to shut down genetically and spatially targeted neurons for
precise time periods. Collection of two photon imaging datasets results in a
whole new set of data analysis problems, due both to differences in fidelity
and temporal resolution in comparison to electrophysiology, and to greater
dimensionality. I will describe how we are adapting information-theoretic
methods for studying the neural code that were developed for singleunit and multi-electrode array data, to an imaging scenario [2,3]. Finally, I
will discuss the prospect of volume-scanning at single cell resolution the
entire depth of the mouse cortical circuit – I will describe how this will be
achievable in the near future, and suggest that the audience think how they
might make use of such a technique.
References
[1] Schultz, SR, Kitamura K, Krupic J, Post-Uiterweer A and Häusser M (2009). Spatial
pattern coding of sensory information by climbing-fiber evoked calcium signals
in networks of neighboring cerebellar Purkinje cells. Journal of Neuroscience,
29:8005-8015.
[2] Panzeri S and Schultz SR (2001). A unified approach to the study of temporal,
correlational and rate coding. Neural Computation, 13(6): 1311-1349.
[3] Schaub MT and Schultz SR (2012). The Ising Decoder: reading out the activity of
large neural ensembles. Journal of Computational Neuroscience. 32(1):101-118.
Royal Academy of Engineering 10
Neural engineering - mentor biographies
Dr Edward Tarte CPhys, FInstP
University of Birmingham
Edward Tarte graduated in 1988 with a BSc (Hons) in Physics from the
University of Bristol. He went on to study for a PhD in the physics of
superconducting devices at the University of Cambridge. He continued
this research between 1992 and 1995 as a Post-Doc in the Cambridge
Materials Science and Metallurgy department. In 1995 he was appointed as
a Senior Assistant in Research in the Cambridge Physics department, where
he began working on Superconducting Quantum Interference Devices
(SQUIDs). In 2000 he was awarded an EPSRC Advanced Fellowship in the
Materials Department in Cambridge. During this period, he investigated
the use of SQUID sensors to detect neuronal activity in-vitro. Edward
moved to Birmingham in 2005 as a University Research Fellow, where his
interest in the detection of bioelectric phenomena developed into a major
part of his research program. His team developed an electrical interface
for the peripheral nervous system, in collaboration with colleagues in
Cambridge and King’s College in London, which has been patented. Other
ways of applying the same technology to bioelectric phenomena are being
developed as well as the use of nanofabrication techniques to improve the
performance of such devices. He was elected a Fellow of the Institute of
Physics in 2008.
Title of presentation
Implantable neural interfaces for prosthesis control
Abstract
In this talk, I will review recent advances in the development of implantable
neural interfaces which allow amputees and paralysed patients to control
prostheses and otherwise interact with their environment using neural
signals. For both groups, the goal of this work is to enable patients to control
devices as naturally and with as many degrees of freedom as possible. For
amputees, very sophisticated robot hands and arms now exist which have
many of the movements of the original limb, but the number of control
signals is limited to two or three using myoelectric electrodes, although
this can be improved by using recent nerve surgical techniques. The use of
electrical signals from the brain to control devices has received considerable
publicity in recent years with Braingate devices being trialled in a number of
paralysed patients, however to maintain long term function, any implanted
device has to withstand the response of the human immune system to its
presence. The performance and even the function of an implanted neural
sensor is determined by the materials from which it is fabricated and its
shape and design. I will discuss the different approaches which have been
adopted for both the brain and peripheral nerve and describe the results
of early stage clinical trials in both cases. I will then focus on the Spiral
Peripheral Nerve Interface (SPNI) developed in a collaboration between the
Universities of Birmingham, Cambridge and King’s College in London, which
highlights the issues involved in the design of a neural interface. The SPNI
is a regenerative electrode array designed to be implanted on the end of a
cut nerve of an amputee. Its design consists of an array of electrodes in the
base of micro-channels whose function is not only to guide nerve fibres,
but also to amplify the neural signal. I will describe the result of in-vivo
tests of this device and the body’s immune response. I will conclude with
a discussion of the potential applications of implantable neural interfaces
beyond prosthesis control, to the provision of sensory and proprioceptive
feedback.
Royal Academy of Engineering 11
Neural engineering - mentor biographies
Professor David Willshaw FRSE, FInstP
University of Edinburgh
David Willshaw is Professor of Computational Neurobiology at the University
of Edinburgh. Since 2001, he has co-ordinated neuroinformatics research in
the UK. Currently, he is the UK Scientific Representative of the International
Neuroinformatics Co-ordinating Facility (INCF), which is a professional
organisation devoted to advancing the field of neuroinformatics, as well as
being Co-ordinator of the UK INCF Node.
He is the grant holder of the Edinburgh Doctoral Training Centre in
Neuroinformatics and Computational Neuroscience funded by EPSRC in
association with MRC and BBSRC.
Since 2002, this Centre has trained over 80 PhD students from the physical
and informational sciences who are applying quantitative approaches to
neuroscience and to neurally-inspired computing.
His research has focussed on neural networks and computational
neuroscience, stretching back over 35 years with over 100 scientific papers.
This has included a spell of combining experimental work with modelling
approaches. He has worked in a variety of research areas including
modelling of (i) associative memory storage in the hippocampus, (ii) the
functioning of basal ganglia, particularly the subthalamic nucleus, and
(iii) the development of patterned nerve connections in the visual and
neuromuscular systems.
He has also developed algorithms for combinatorial optimisation (with
Richard Durbin he developed the Elastic Net algorithm for the Travelling
Salesman Problem). In his current research, he leads a multi-centre
Wellcome Trust funded project combining modelling approaches to
investigate the roles of neural activity and molecular signalling in the
formation of ordered nerve connections in the mouse retinocollicular
system.
Since 1984, he has held long term research support from the UK Medical
Research Council and the Wellcome Trust. He was the recipient of the IEEE
Neural Networks Council Pioneer Award in 1992. From 1999-2005, he
was Editor-in-Chief of the computational neuroscience journal Network:
Computation in Neural Systems. With colleagues, he has just published a
Cambridge University Press text book Principles of Computational Modelling
in Neuroscience.
Title of presentation
Neuroinformatics
Abstract
The brain is the most complex and mysterious system known to mankind.
To understand how the brain works not only will be a triumph in itself but
also will improve massively our health by preventing and combating disease.
We need to understand the functioning of the billions of nerve cells – the
computing units of the brain. We need to know how they are formed from
the gene, how they interconnect and how they influence one another.
Royal Academy of Engineering 12
Neural engineering - mentor biographies
An enormous amount of data already exists about all these aspects of the
brain. This ranges from how the programs in the genes influence nerve
cell function, how nerve cells communicate with one another through
exchanging electrical signals and how different parts of the brain are active
when we carry out specific tasks. But despite this the brain is still very
poorly understood. The hope is that we can use all these data to understand
how to relate the functioning of the genes coupled with the influences from
the world around us gives rise to our thoughts and our actions.
Neuroinformatics offers new methods, using techniques from physics,
engineering, mathematics and computer science, for making sense of this
vast amount of data. To do this we have to devise new ways to extract the
vast amount of information from the data to be analysed and make sense of,
as well as to share the expensive data. This process will also involve making
and testing computer models of how we think that a particular part of the
brain functions, which can be compared with the actual data and can then
be used to predict the results of further experiments.
In this talk I will review progress in the field of neuroinformatics. I will
also introduce the achievements of the International Neuroinformatics
Coordinating Facility (INCF), which was formed in 2005 to coordinate
neuroinformatics worldwide.
Royal Academy of Engineering 13
Young researchers
abstracts
Royal Academy of Engineering 14
Neural engineering - young researchers abstracts
Richard Barrett1, Edward J. Tarte1, James J. FitzGerald2, Natalia Lago3, Samia Benmerah1,
Jordi Serra3, Christopher P. Watling4, Ruth E. Cameron4, Stéphanie P. Lacour5,
Stephen B. McMahon3 and James W. Fawcett2
As presented by Richard Barrett, University of Birmingham
1
2
3
4
5
School of Electronic, Electrical and Computer Engineering, University of Birmingham, Birmingham, UK
Centre for Brain Repair, University of Cambridge, Cambridge, UK
Neurorestoration Group, King’s College London, London, UK
Department of Materials Science and Metallurgy, University of Cambridge, Cambridge, UK
Institute of Microengineering, EPFL, Lausanne, Switzerland
EPSRC and Basic Technology Research Programme
Spiral peripheral nerve interface: a novel interface design
Introduction
We have developed a novel peripheral nerve interface device based on photosensitive polyimides in which nerve
fibres regenerate through channels containing electrodes. This design has advantages over others in terms of both
recording and stimulation. Initial tests of the device carried out in-vivo show successful regeneration and have
demonstrated that they can record signals from regenerated axons. These results, which were obtained with the
first design for our device, enabled us to prove the microchannel interface concept. However we found that there
were problems in establishing robust electrical connections to these implants. We have revised the fabrication
process to allow the whole structure to be fabricated in photosensitive PSPI, which enables a through hole rivet
bonding process to be used. This talk will describe the fabrication and performance of the novel devices produced
using both processing routes and to discuss biocompatibility issues.
Method
These Spiral Peripheral Nerve Interfaces (SPNIs) are fabricated using well established photolithographic techniques
developed in the semiconductor industry. Through spin coating and curing, layers of polyimide can be built up into
complicated three-dimensional structures. On top of a silicon carrier wafer, with a sacrificial layer of Poly(methyl
methacrylate) PMMA, photosensitive polyimide (PSPI) is defined into a substrate layer. Further layers of polyimide
are spun on top of the substrate to produce 100micron wide channels in a 100micron thick PSPI. Metal wiring is
integrated between the two and encapsulated using a thinner PSPI layer. The PMMA serves as a sacrificial material
which is wet etched using a mixture of MIBK and Acetone to release the devices from the carrier wafer. Once
released, the devices are rolled and inserted into a silicone tube, ready for implantation. The latest design has been
mechanically tested using a pull tester and the impedance has been measured in-vitro to serve as a comparison
with the earlier designs.
Results
Initial tests of the device carried out in-vivo show successful regeneration and have demonstrated that we
can record signals from regenerated axons. These results, which were obtained with the first design for our
device, enabled us to prove the microchannel interface concept. However we found that there were problems in
establishing robust electrical connections to these implants. To alleviate these issues the fabrication process was
modified to incorporate PSPI in the substrate layers, allowing for the rivot bonding process. Careful consideration
is made to the bending stresses within the device during the rolling process, requiring further modifications to the
design process. The latest device has been shown to be a viable recording interface with impedance in the range of
300k Ohms to 1M Ohms @ 1kHz, and has also been shown to be a mechanically stable structure for implantation.
Discussion
The concept of the SPNI has been proven through the original device, however to create a feasible peripheral nerve
implant the method of generating stable electrical connections had to be improved. A stable connection platform
would allow an increase in the number of channels to interface with (currently 20) which lead to an increase in the
potential degrees of freedom available to control a prosthetic device. This would also require multiplexing, which
was impossible using the original design but could be incorporated into the updated SPNI device with further work.
References
Benmerah, S., Lacour, S.P., Tarte, E., 2009. Design and Fabrication of Neural Implant with Thick Microchannels based on Flexible
Polymeric Materials. 31st Ann Int Conf of the IEEE EMBS.
FitzGerald, J., Lago, N., Benmerah, S., Serra, J. Watling, C. P. Cameron, R.E., Tarte, E., Lacour, S.P., McMahon, S. P., Fawcett, J.W. 2012.
A regenerative microchannel neural interface for recording from and stimulating peripheral axons in vivo. J. Neural Eng. 9 016010
Royal Academy of Engineering 15
Neural engineering - young researchers abstracts
Deren Barsakcioglu
Centre for Bio-Inspired Technology, Imperial College London
deren.barsakcioglu10@imperial.ac.uk
Funded by EPSRC
Ultra-low-power real-time spike sorter for neural prostheses
Introduction
Neuroprosthetics aim to restore lost sensory and motor function of millions of people by interfacing directly with
their nervous system. Brain-based neuroprosthetic interfaces achieve this using microelectrode arrays that record
multiple neuronal spiking activities close to each electrode.
To identify which of these neurons have fired, a signal processing step called spike sorting is used. With trends in
electrode technology thousands of channels can now be recorded. It is beyond current bandwidth capabilities to
transmit raw data of this magnitude for totally implantable systems. Thus spike sorting is a vital step in reducing
transmitted data while maintaining quality of information transmitted.
Method
We propose to use template matching as our spike sorting method and develop a fully autonomous resource
efficient real-time system. Within this context, the first step in the project is to identify and establish the accuracy
and complexity trade-off of key parameters on spike sorting. This includes testing several signal processing
techniques to improve accuracy, while estimating the required hardware resources (memory used, number
of operations, power consumption etc.). In the following stages, novel algorithms will be developed to ensure
automatic and adaptive operation (classification and calibration). Once all trade-offs are established and algorithms
are selected, the ways in which these algorithms can be most efficiently implemented in hardware will be
investigated.
Results
As the first step in my research, the work to date has investigated the parameters associated with template
matching and the analogue-front-end preceding the spike sorter. Having identified these parameters, they were
quantified in terms of spike sorting accuracy as well as complexity (power and area). After establishing the key
hardware trade-offs, the next stage in the project is to develop the spike sorting algorithms and implement them in
CMOS.
Discussion
My work is part of a project to develop a fully integrated implantable neural interface with a real-time spike sorter,
minimising power consumption and area while maintaining the high performance achieved by off-line equivalents.
In order to overcome power-bandwidth bottleneck due to ever increasing number of channels recorded, on-chip
spike sorting is essential. Achieving such an on-chip real-time resource efficient and self-calibrating system will
provide neural prosthetics and neuroscientists with more data to utilize than ever before.
Royal Academy of Engineering 16
Neural engineering - young researchers abstracts
Lionel Chaudet1, Prof Mark Neil1, Dr Patrick Degenaar2, Dr Kamyar Mehran2,
Dr Rolando Berlinguer-Palmini2, Prof Brian Corbet3, Pleun Maaskant3,
Mr David Rogerson4, Dr Peter Lanigan4, Prof Ernst Bamberg5, and Dr Botond Roska6
As presented by Lionel Chaudet, Imperial College London
1
2
3
4
5
6
Department of Physics, Photonics Group, Blackett Lab, Imperial College, London, UK
School of Electrical, Electronic and Computer Engineering, Newcastle University, Newcastle upon Tyne, UK
National Tyndall Institute, Cork, Ireland
Scientifica, East Sussex, UK
Department of Biophysical Chemistry, Max Planck Institute of Biophysics, Frankfurt, Germany
Friederich Miescher Institut, Basel, Switzerland.
lionel.chaudet09@imperial.ac.uk
Seventh Framework Programme (FP7)
Improving opto-electronic neural stimulation with micro-LED arrays and optics
Introduction
The breakthrough discovery of a nanoscale optically gated ion channel protein, Channelrhodopsin 2 [1], allowed
neuron cells to be made photosensitive through genetic re-engineering. Combined with a genetically expressed ion
pump, Halorhodopsin, it has become possible to directly stimulate and inhibit individual action potentials with light
alone. This work reports developments undertaken as part of the European project, OptoNeuro, which is developing
ultra-bright electronically controlled optical array sources with enhanced light gated ion channels and pumps for use
in systems to further our understanding of both brain and visual function.
Method
Micro-LED arrays permit spatio-temporal control of neuron stimulation on sub-millisecond timescales. However
they are disadvantaged by their broad light emission distribution and low fill factor. We present the design and
implementation of a projection and micro-optics system for use with a micro-LED array consisting of a 16x16 matrix
of 25 μm diameter micro-LEDs with 150 μm centre-to-centre spacing. Its emission spectrum is centred at 470
nm overlapping the peak sensitivity of ChR2. The projection system images the array output onto micro-optics to
improve the fill-factor by capturing a larger fraction of the LED emission and directing it correctly to the sample
plane.
Results
The fill-factor was increased on the microscope plane from 2 % without projection and micro-optics (see Fig. 1(a))
to nearly 100 % with it (see Fig. 1(b)). The entire projection system is integrated into a microscope prototype to
provide stimulation spots at the same size as the neuron cell body (~10 μm). The collection efficiency was increased
from 0.13 % to 1.76 %.
Figure 1: Illumination on the sample (microscope plane) with 10x10 micro-LEDs
(a) without Projection and micro-optics and (b) with it
Royal Academy of Engineering 17
Neural engineering - young researchers abstracts
Discussion
The results obtained corresponded to the ones simulated in the optical software Zemax [2] (see Fig. 2 (a) and (b))
and the theory with an expected collection efficiency of 1.56%. This approach allows low fill factor arrays to be used
effectively, which in turn has benefits in terms of thermal management and electrical drive from CMOS backplane
electronics.
Figure 2: Simulation of the illumination with 10x10 micro-LEDs on the sample using the optical software “Zemax” with projection
and micro-optics on (a) a full-field spot diagram and (b) a spot diagram of each of those positions
References
[1] Buchen, L., 2010. “Illuminating the brain”, Nature, Vol. 465, Issue 7294, Page 26-28
[2] Radiant Zemax, “http://www.radiantzemax.com/en”
Royal Academy of Engineering 18
Neural engineering - young researchers abstracts
Graeme Coapes1, Terrence Mak2 and Alex Yakovlev1
As presented by Graeme Coapes, Newcastle University
1 School of EEE, Newcastle University, NE1 7RU
2 Department of Computer Science and Engineering, Chinese University of Hong Kong
+44 (0) 191 222 7340 graeme.coapes@ncl.ac.uk
Funded by EPSRC
Silicon neuron node capable of large-scale simulations of bio-physically
realistic ion channels
Introduction
The latest developments within brain-machine interfaces allows for the direct recording and stimulation of single
neurons. This technology allows for the creation of hybrid bio-silicon systems, whereby silicon models of neurons
are able to interact with biological elements in a real-time closed-loop environment. This offers the opportunity to
not only study further the operation of the brain, but also to develop novel neural prosthetic designs, which can
repair or restore the functionality of brain regions.
Typically neuronal models for hybrid systems are created using software and standard desktop computers.
However, due to the underlying sequential architecture of such systems the size and performance of these
models are limited. A dedicated hardware approach is also becoming popular due to the capabilities and potential
that is offered, but this approach suffers in development costs, in terms of finance, time and complexity. Using
reconfigurable hardware, such as field-programmable gate arrays (FPGAs), however reduces these costs
significantly and provides comparable levels of performance.
Design
The developed silicon neuron node allows for the simulation of over 50,000 bio-physically realistic Hodgkin-Huxley
neurons in real-time. The neurons to be simulated can be programmatically defined, in a similar manner to writing
software-based simulations.
The concept of the node’s design is inspired by typical microprocessor architecture, but with the standard datapath
replaced by a dedicated and optimized arithmetic block which can update a neuron’s output within under 200ns.
Although similar times may be achieved using a standard desktop processor, the FPGA-based design allows for
pipelining of the calculations, improving the number of parallel computations and providing substantial performance
benefits.
Discussion
The design of the neuron node is the first-step in developing a large-scale architecture that is capable of real-time
closed-loop interactions between large populations of bio and silicon elements. We have previously shown the
potential performance benefits of this node, in particular the optimized scalability, in terms of resource usage and
model accuracy.
Future work will include the development of a complete system, which will involve the design of an optimized
silicon synaptic unit and a high-performance network and communication architecture suitable for bio-realistic
simulations.
Figure 1 – Closed-Loop Hybrid Network
Table 1 – Comparison of Design Approaches
References
Coapes G. et al., 2012. A Scalable FPGA-based Design for Field Programmable Large-Scale Ion Channel Simulations. 22nd
International Conference on Field Programmable Logic and Applications, In Press
Royal Academy of Engineering 19
Neural engineering - young researchers abstracts
Simon Davies
International Digital Laboratory, University of Warwick
International Digital Laboratory, University of Warwick, CV4 7AL
davies_s@wmg.warwick.ac.uk
SKB Scholarship, Warwick Manufacturing Group
Empirical mode decomposition for feature extraction in P300 responses for BCI
Introduction
A P300 is an evoked potential generated by brain activity in response to the user receiving an “oddball” stimulus [1].
This is when an expected but rarely occurring change is detected by the user’s sensory inputs, e.g. hearing a lowfrequency tone after a succession of high-frequency tones. The P300 is so-called because it generates a positive
change in potential voltage difference on average 300ms after the stimulus is received. We can measure the P300
response using an electroencephalogram (EEG). However, EEG signals suffer from a high Signal-to-Noise ratio
due to their low power and multiple underlying sources. Raw data from an EEG must be processed before it can be
accurately detected if a P300 occurred or not. In this case Empirical Mode Decomposition (EMD) [2] [3], an iterative
and empirical method that breaks a signal down into a group of harmonics and residual noise, was evaluated as a
possible method for feature extraction.
Method
The method was applied to Data Set II from BCI Competition III [4]. Data Set II consists of P300 training and test data
for two subjects. Each subject focused on a 6x6 character grid. Each row and column flashed once in random order.
This process then repeated 15 times. The data from these 180 epochs was used to identify the character the user
was focusing on. This was done for 100 characters. Using all of the training data an idealised set of Intrinsic Mode
Functions (IMFs) was constructed for both the P300 and non-P300 signal. Normalised FFTs of each of the ideal IMFs
were recorded.
Next, EMD was applied to each set of character training data and normalised FFTs calculated from each IMF.
The Euclidean distance between each FFT and the ideal P300 FFTs was measured. This was then normalised
and adjusted to see how proportionately close each IMF was to the ideal IMFs. The training data IMFs were then
weighted and summed to form the processed signal. The same process was repeated with the non-P300 FFTs and
the two data sets concatenated together. An FFT of this data is then generated and the first 60 samples (0-12Hz)
are stored. This is done to all the training data epochs and is used to train a Support Vector Machine (SVM). The test
data goes through the same process and is classified using the SVM created.
Results
Currently performance stands at 66% sensitivity and 65% specificity.
Discussion
EMD is a useful method as it can be applied to any non-stationary data, making it suited to EEG. Few assumptions
are needed to be made about the data, and no specific bases need to be assumed (as would be the case for wavelet
transforms, say). Here we are using the FFT as a means of characterising the IMFs followed by the Euclidean
distance measure to assess closeness, this choice is being assessed as the FFTs do not capture the wave shapes
which we think are necessary for the P300 detection, similarly the Euclidean distance may not be appropriate.
These are current areas of research in this application.
References
[1] Nijboer F., Sellers E., Mellinger J., Jordan M., Matuz T., Furdea A., Halder S., Mochty U., Krusienski D., Vaughan T., Wolpaw J.,
Birbaumer N., Kubler A., 2008, “A P300-Based Brain–Computer Interface for People with Amyotrophic Lateral Sclerosis”,
Clinical Neurophysiology, Vol. 119, pp1909-1916
[2] Huang N., Shen Z., Long S., Wu M., Shih H., Zheng Q., Yen N., Tung C., Liu H., 1998, “The empirical mode decomposition and the
Hilbert spectrum for nonlinear and non-stationary time series analysis”, Proceedings of the Royal Society A, Vol. 454, pp903905
[3] Solis-Escalante T., Gentiletti G., Yanez-Suarez O., 2006, “Single Trial P300 detection based on the Empirical Mode
Decomposition”, Proceedings of the 28th IEEE EMBS Annual International Conference, pp1157-1160.
[4] Wolpaw J., Schalk G., Krusienski D., BCI Competition III Data Set II, http://www.bbci.de/competition/iii/#data_set_ii
Royal Academy of Engineering 20
Neural engineering - young researchers abstracts
Vaibhav Gandhi, G. Prasad, D. Coyle, L. Behera and T. McGinnity
As presented by Vaibhav Gandhi, University of Ulster
Intelligent Systems Research Centre, University of Ulster, BT48 7JL
+44 (0) 28716 75392 gandhi-v@email.ulster.ac.uk
Funder: UKIERI grant ‘Innovations in Intelligent Assistive Robotics’, InvestNI and the Northern Ireland Integrated
Development Fund
EEG filtering with quantum neural networks for a Brain-Computer Interface (BCI)
Introduction
Electroencephalogram (EEG) recorded during motor imagery (MI) based communication using a Brain-computer
interface (BCI) is inherently embedded with non-Gaussian noise while the actual noise-free EEG has so far been
elusive. This paper presents a novel neural information processing architecture which involves deploying the
Schrodinger Wave Equation (SWE) to filter noise from EEG.
Method
The proposed Recurrent Quantum Neural Network (RQNN) (cf. Figure 1) represents a non-stationary stochastic
signal as time varying wave packets [1]. The basic approach is to ensure that the statistical behaviour of the input
signal is properly transferred to the wave packet associated with the response of the quantum dynamics of the
network. At every computational sampling instant, the EEG signal is encoded as a wave packet which can be
interpreted as the probability density function (pdf) of the signal at that instant. The features in the form of Hjorth
parameters and band-power are then extracted from the RQNN filtered EEG. These features are classified using
linear discriminant analysis (LDA).
Figure 1: RQNN model
Results
The RQNN is demonstrated to effectively filter noise from simple signals such as DC, staircase DC and sinusoidal
with model parameters optimized using particle swarm optimization (PSO). The RQNN is significantly better than
the Kalman model while filtering the DC signal. A two-step inner-outer 5-fold cross-validation approach is utilized
for selecting the RQNN model parameters to suit individual subjects. The filtered signals from the subject specific
RQNN model during the training and the evaluation stages are shown to be more separable than the raw EEG,
Savitzky-Golay filtered EEG or raw EEG with the power spectral density or the Bispectrum based features for the BCI
competition IV 2b dataset.
References
[1] V. Gandhi, V. Arora, L. Behera, G. Prasad, D. Coyle and T. McGinnity, 2011. “EEG denoising with a recurrent quantum neural
network for a brain-computer interface,” Neural Networks (IJCNN), The 2011 Inernational Joint Conference on, pp 1583-1590
Royal Academy of Engineering 21
Neural engineering - young researchers abstracts
Julian H. George, Hua (Cathy) Ye and Zhangfeng Cui
As presented by Julian George, University of Oxford
Institute of Biomedical Engineering, Department of Engineering Science, University of Oxford, OX3 7DQ
+44 (0) 1865 617694 julian.george@eng.ox.ac.uk
Funded by BBSRC: BB/H008608/1
One network to another - engineering a neural interface through
micro-porous membranes
Introduction
To facilitate the creation of an interface neural network layer, we investigated the use of microporous membranes
to control neurite outgrowth between networks.
Methods
We investigated the effect of membrane coating (laminin, collagen type I and poly-D-lysine) and pore diameter
(0.4μm to 5μm diameter) on the neurite outgrowth of retinoic acid treated NT2-D1 and SH-SY5Y cells, through
commercially available track-etched polycarbonate Isopore® membranes (see fig 1), similar to those used in neurite
outgrowth assays [1]. Neurite extension towards a second neural network seeded below the membranes was also
investigated. After 5-14 days in culture, neurons were fixed and immunostained or prepared for SEM microscopy.
Figure 1: Porous membranes, secured using PDMS rings in 24 well plates (A, B), were used to culture neurons
and study neurite outgrowth for single (C), and double layer networks (D) and towards encapsulated networks (E).
An SEM image reveals neurite outgrowth through 5μm diameter pores (F).
Results and discussion
Pore diameters of 0.8μm or smaller were found to restrict neurite outgrowth, whilst pore diameters larger than
3μm permitted cell migration through the membranes. For pore diameters larger than 1.2μm, multiple neurites
extended through each pore. Membrane coating was found to have an effect on neurite fasciculation.
The advantage of culturing an interface neural network through a porous membrane is that the membrane
facilitates the controlled manipulation of the interface, without permitting mixing of the two cell populations.
This approach may also be useful in-vivo, when it is desirable to keep implanted cells separated from host cells [2].
Through use of established techniques, such as the selective patterning of ligands and the modification of surface
topography, it should be possible to use porous membranes to structure the interface network and align cells with
the transducers of the interface array.
References
[1] Smit M, Leng J, Klemke RL. 2003. Assay for neurite outgrowth quantification. Biotechniques.35(2):254-6.
[2] Yamazoe H, Keino-Masu K, Masu M. 2010. Combining the cell-encapsulation technique and axon guidance for cell
transplantation therapy. J Biomater Sci Polym Ed. 21(13):1815-26.
Royal Academy of Engineering 22
Neural engineering - young researchers abstracts
Vasiliki Giagka*, N. Donaldson** and A. Demosthenous*
As presented by Vasiliki Giagka, University College London
* Analogue and Biomedical Electronics Group, Department of Electronic & Electrical Engineering,
University College London, Torrington Place, London WC1E 7JE
** Implanted Devices group, Department of Medical Physics and Bioengineering, University College London,
+44 (0)20 7679 4155 v.giagka@ee.ucl.ac.uk
This work is part of the European Research project NEUWalk, funded by the European Community’s Seventh
Framework Programme [FP7/2007-2013] under grant agreement no. CP-IP 258654.
An implantable stimulator system for neuro-rehabilitation in paralyzed rats
Introduction
Recent studies in humans and rats [1] suggested epidural spinal-cord electrical stimulation as a means to facilitate
locomotor recovery after spinal-cord injury. In order to fully understand the mechanism involved we are developing
an implantable system, comprised of a flexible epidural electrode array and application specific stimulating
electronics, to be used in rats. The development of such a system is challenging due to the tight area and power
constraints posed by the small dimensions of the rats. It needs to be suitable and robust enough for chronic
implantation and provide high flexibility in terms of stimulation sites.
Method
In the first generation of the system a passive electrode array was used together with an external stimulator.
The electrode array was fabricated using silicone rubber and laser-patterned platinum foil, while stainless steel wires
were soldered on the platinum tracks to interconnect the array to the external stimulator. Chronic implantation [2]
revealed the need to minimize the tracks that connect the electrodes to the electronics, as they constitute the main
cause of implant failure. To achieve that, a demultiplexing integrated circuit (IC) that communicates via only two wires
[3]
is being designed to be embedded on the array and drive the electrodes in any possible configuration.
Results
Results of the implantation of the first version of the system, as well as simulation results of the basic building
blocks of the demultiplexing IC will be presented.
Discussion
The IC under development is designed to act as a demultiplexer, redirecting the stimulus current to the selected
electrodes. To better distribute the dissipated heat we are splitting this IC into three identical parts, each of which
drives four electrodes and is embedded on the twelve-electrode array. The connection between the IC and the
electrode array needs to be investigated.
References
[1] Van den Brand R. et al., May 2012, “Restoring Voluntary Control of Locomotion after Paralyzing Spinal Cord Injury”, Science,
vol. 336, no. 6085, pp. 1182-1185.
[2] Giagka V., Vanhoestenberghe A., Wenger N., Musienko P., Donaldson N., and Demosthenous A., April 2012, “Flexible platinum
electrode arrays for epidural spinal cord stimulation in paralyzed rats: An in vivo and in vitro evaluation,”, in proceedings, 3rd
Annual Conf. IFESSUKI 2012, pp. 52-53.
[3] Giagka V., Demosthenous A., and Donaldson N., June 2012, “Towards a Low-Power Active Epidural Spinal Cord Array Controlled
Through a Two Wire Interface,” in proceedings, PRIME 2012, Aachen, Germany.
Royal Academy of Engineering 23
Neural engineering - young researchers abstracts
Simao Laranjeira, P. Orlowski, T. Lillicrap, C. Lueck, A. Neely and S. Payne
As presented by Simao Laranjeira, University of Oxford
Department of Engineering Science, University of Oxford, Old road campus, Hedington, OX3 7UQ
+44 (0) 7786546215 simao.laranjeira@eng.ox.ac.uk
Modelling glutamate dynamics and associated tissue temperature change in
brain cells after stroke
Introduction
Identification of salvageable brain tissue is a major challenge for the planning of the treatment of stroke
patients. We investigate the role of brain temperature as a biomarker for guiding decisions. Imaging of the ischaemic
human brain with quantitative temperature MRI shows an increase of temperature in the ischaemic tissue [1].
We hypothesise that the excessive heat generated is due to accelerated metabolism and particularly pronounced
after uncontrolled release of glutamate, which in excessive concentration is toxic for neurons [2, 3]. Establishment of
a quantitative relationship between perfusion and the change of temperature and glutamate concentration would
permit validation of the aforementioned hypothesis.
Method
To achieve this goal we are expanding on the model from [4] of brain metabolism. Release of glutamate was included
as described in [5]. Temperature of the system was modelled to be dependent on arterial blood temperature,
metabolic heat rate and the initial temperature difference between arterial blood and brain tissue. Arterial blood
temperature is kept constant and independent of blood perfusion. Metabolic heat rate was modelled proportional to
the consumption of ATP by the Na2+/K+ pump. At steady state, blood removes the heat produced by the metabolic
rate and the system is kept at 310.15K. From [6] brain tissue can have a temperature 0.2K to 0.6K higher than blood.
The effect of temperature change on reaction rates was modelled using the Q10 rule.
Results
In our experiments we investigate the effect of glutamate release on the systems temperature in two cases:
1) when the level of cerebral blood flow (CBF) is changed between 0.2 and 0.8; 2) when glutamate is artificially
triggered at a healthy CBF level (above 0.8). In both cases the difference between brain and blood temperature
was kept at 0.4K. In the first scenario, when reducing CBF by 80% to simulate a stroke, there was an increase in
temperature of 0.02K, during the first 100 seconds. After that, temperature decreased until it reached a steady
state of 310.3K. In the second scenario the full depletion of glutamate was responsible for an increase of 0.14K.
Even in the extreme case of a temperature difference between brain tissue and blood of 0.6K the increase of
temperature in the second scenario was 0.14K.
Discussion
Our own MRI temperature measurements of ischaemic brains and the results from [1] show that there is an increase
of 1 ± 1K in the brain tissue suffering from ischaemia. The result presented previously showed that the effect
of glutamate is only a fraction of the prediction. However, it was found that when there is no blood supply and
all the metabolic agents are consumed there could only be an increase of 0.261K from metabolic heat release.
Therefore, our results are well within bounds of the system and an increase of 1K without any blood supply due to
the mechanisms here described is not possible. In the near future the following extensions for the model are being
considered: 1) the effects of perfusion on arterial temperature and volume; 2) developing a more sophisticated
model for metabolic heat removal; 3) incorporate cerebral tissue heat diffusion. Also, as future work we want to
investigate if glutamate release is the sole agent responsible for the 0.2K difference.
References
[1] KARASZEWSKI, B., WARDLAW, J., MARSHALL, I., CVORO, V. and WARTOLOWSKA, K., 2006. Measurement of brain temperature
with magnetic resonance spectroscopy in acute ischemic stroke. Annals of Neurology, 60(4), pp. 438-446.
[2] CLOUTIER, M., BOLGER, F., LOWRY, J. and WELLSTEAD, P., 2009. An integrative dynamic model of brain energy metabolism
using in vivo neurochemical measurements. Journal of computational neuroscience, 27(3), pp. 391-414.
[3] SHULMAN, SHULMAN, R.G. and ROTHMAN, D.L., 2004. Brain Energetics and Neuronal Activity: Applications to fMRI and
Medicine. Chichester: John Wiley Sons, Ltd, pp. 29-3 , 71-64.
[4] ORLOWSKI, P., CHAPPELL, M., PARK, C., GRAU, V. and PAYNE, S., 2011. Modeling of pH dynamics in brain cells after stroke.
Interface Focus, 1(3), pp. 408-16.
[5] SUN, J., PANG, Z.P., QIN, D., FAHIM, A.T. and ADACHI, R., 2007. A dual-Ca 2 -sensor model for neurotransmitter, release in a
central synapse. Nature, 450(7170), pp. 676-82.
[6] HAYASHI, N, DIETRICH, D. W., 2003, Brain Hypothermia Treatment, Springer, pp. 57
Royal Academy of Engineering 24
Neural engineering - young researchers abstracts
Jun Wen Luo1, Terrence Mak1, Peter Andras2 and Alex Yakovlev1
As presented by Jun Wen Luo, Newcastle University
1 School of Electronic, Electrical Engineering
2 School of Computer Science
j.w.luo@ncl.ac.uk
A novel hardware architecture for large scale hybrid bio-silicon network
Introduction
Hybrid network (dynamic clamp) is a novel technique that bridges the communication gap between neurological
and artificial systems. However, software implementation imposes a limited computational capability, which can
introduce restriction in system reliability and scalability, especially when biophysical means are given to neuronal
models. One such technology to overcome these problems is analogy VLSI circuits; however the relatively long
design and fixed architectures can be a bottleneck in developing a new application. Alternatively supercomputers
are employed to simulate large-scale neuron models, the lack of portability and complex configuration hinder the
performance of these devices in practical hybrid applications. FPGA is another solution for solving this problem,
limited hardware and extremely update periods are two drawbacks for implementation of hybrid network.
We proposed a novel hardware architecture based on Time Division Multiplexer technique for hybrid network.
Firstly a complete biological Centre Pattern Generator (CPG) model with 14 neurons is implemented, and this
model is interacting with imperfect real biological network to rehabilitate its standard rhythm patterns. Results
demonstrated system accessibility and reliability.[1, 2]
Method
The system architecture is below:
Results
Compare to the real control pyloric rhythm, simulation rhythms mainly followed real rhythm phase relationships PD –
LP – PY. For dynamic clamp results, the hybrid system is able to restore disable biological neuron patterns.
Discussion
Results indicate FPGA is a good platform for large scale hybrid network. However, complicated hardware
architecture and design flow may constrain its common applications.
References
[1] Luo J,W,M. T. Yu B, Andras P, Yakovlev A., “Towards neuro-silicon interface using reconfigurable dynamic clamping,” Conf Proc
IEEE Eng Med Biol Soc, pp. 6389-92, 2011.
[2] Graeme Coapes,T. M. Jun Wen Luo, Alex Yakovlev, Chi- Sang Poon, “A Scalable FPGA -based Design for Field Programmable
Large -Scale Ion Channels Simulations,” Conference on Field Programmable Logic and Applications, 2012(in press).
Royal Academy of Engineering 25
Neural engineering - young researchers abstracts
James McGuinness and B. P. Graham
As presented by James McGuinness, University of Stirling
Institute of Computing and Mathematics, School of Natural Sciences, University of Stirling, FK9 4LA
jmc@cs.stir.ac.uk
Role and implications of population heterogeneity on vestibular ocular reflex
response fidelity
Introduction
The Vestibular-Ocular Reflex (VOR) is a sensori-motor reflex characterised by high fidelity, low latency
compensatory eye movements in response to head movements. A major component of the VOR is the linear,
rotational reflex, which produces eye movements in order to compensate for angular head rotations.
This component of the VOR is produced through an elementary 3 neuron arc. However, single cell responses of
the Vestibular Nucleus neurons involved show non-linearity and distortion for high frequency inputs suggesting,
instead, that the required signal transmission is achieved by response of a population of neurons. In addition,
evidence for a functional role of population variance and heterogeneity in sensori-motor systems is currently
emerging.
Method
Using previously published, biophysically complete, compartmental models of Medial Vestibular Nucleus neurons,
we investigate the affect of variance and Heterogeneity of ion channel density on the fidelity of the population
response to a broad range of input frequencies. In addition, various commonly used methods for the analysis
and estimation of the population response will be discussed in regards to their suitability, accuracy and inherent
limitations.
Results
It will be shown that Heterogeneous populations (in which the extent of certain conductances varies for each
member) produces a response with greater fidelity than Homogeneous populations (in which these conductances
do not vary).
Discussion
The functional implications of population variance and Heterogeneity will be discussed in the context of the VOR,
along with other emerging evidence for a functional role for population Heterogeneity. In addition, the techniques
used for analysis of population responses will be discussed.
Royal Academy of Engineering 26
Neural engineering - young researchers abstracts
Sivylla-Eleni Paraskevopoulou
Centre of Bio-Inspired Technology, Imperial College London
s.paraskevopoulou09@imperial.ac.uk
Funded by EPSRC
Ultra-low power implantable platform for next generation neural interfaces
Introduction
My PhD thesis is part of an inter-university multi-disciplinary project aiming to develop a novel multi-channel
implantable neural interface for the recording and sorting of single neuron activity in the brain. Our group’s
contribution is the implementation of the ultra low-power mixed signal circuitry while the University of Leicester
is developing the offline spike sorting calibration algorithms and the Institute of Neuroscience at Newcastle the
implementation of the system in animal models.
Method
My work on this project focuses on the analogue front-end of the neural interface. The front-end is responsible for
conditioning, amplifying and filtering the recorded signal, and performing initial signal processing. The two main
questions guiding my research are: (1) given the communication bandwidth restrictions and power limitations due
to heat dissipation, which can cause brain tissue damage, how much can the channel count be increased using
ultra low power electronics, and (2) how much of the signal processing can be integrated in the front-end without
sacrificing the performance accuracy of the algorithms.
Results
In the scope of my research, so far, the key accomplishments are that we have investigated, from a system level
perspective, the various possible configurations of neural interfaces, how transmitted information is represented
in each scheme, what level of processing is required, the transmission bandwidth requirements for each channel,
and how power consumption may be optimized. Moreover, we have designed a system that locates the spike peaks,
encodes spike timing information, computes the spike rate and holds the spike peak amplitude value. Finally, a novel
neural front-end has been implemented with tunable bandwidth and featuring automatic gain control that enables
reduction of the ADC’s resolution and uncalibrated monitoring of multiple channels.
Discussion
The ultimate goal of this research is the hardware implementation of a fully-automated channel encompassing
state-of-the art ultra-low power circuitry that will handle the recorded neural signal and produce high-level data.
References
A. Eftekhar, S. Paraskevopoulou, and T. Constandinou, “Towards a next generation neural interface: Optimizing power, bandwidth
and data quality,” IEEE BioCAS, pp. 122-125, 2010.
S. Paraskevopoulou, and T. Constandinou, “A sub-1μW neural spike-peak detection and spike-count rate encoding circuit,” IEEE
BioCAS, pp. 29-32, 2011.
S. Paraskevopoulou, and T. Constandinou, “An Ultra-Low-Power Front-End Neural Interface with Automatic Gain for Uncalibrated
Monitoring,” IEEE ISCAS, 2012.
Royal Academy of Engineering 27
Neural engineering - young researchers abstracts
Arsam Shiraz*, A. Demosthenous and A. Vanhoestenberghe
As presented by Arsam Shiraz, University College London
* Electronic & Electrical Engineering Department, UCL, Torrington Place, London WC1E 7JE
+44 (0) 20 7679 4155 a.shiraz@ucl.ac.uk
Design and development of a closed-loop active neuromodulation device for
treating urinary incontinence
Introduction
Urinary incontinence (UI) is the involuntary passing of urine. This socially restrictive condition is consistently
associated with adverse effects on quality of life for patients. The UI can be the outcome of a range of factors
including pelvic trauma, neurological trauma or disease, and cognitive impairment. The inefficiency of the existing
therapeutic techniques, intense side effects of the prescribed drugs, and invasive and costly nature of the surgeries
show the need for the design and development of new forms of treatment.
It has been shown that trans-rectal stimulation of the pudendal nerve upon sensing an adequate
electromyographic (EMG) signal from the external sphincter can effectively inhibit detrusor hyper-reflexia in spinal
cord injury patients with neurogenic bladder [1]. An active device [2] is being developed which comprises of a silicone
rubber anal probe and a battery powered silicon chip capable of stimulation and EMG recording packaged inside of
the probe.
Method
A model of electrode-tissue interface should be developed to be used in simulations. Furthermore, voltage
compliance and other electrical specifications should be derived to be able to design the electronic circuits. The
passive probe shown in figure 1 was inserted in the anus of a male volunteer and 200 μs current pulses repeated at
a frequency of 15 Hz with varying amplitudes (up to 45 mA) were delivered through the front electrodes while the
voltage between the electrodes was monitored. In another setting, pulses of different amplitudes and durations
were delivered and the response of the external sphincter was monitored via a pressure sensor to construct the
strength-duration curves for the trans-rectal stimulation of the pudendal nerve for different repetition frequencies.
Figure 1: Passive anal probe for pudendal nerve stimulation
and external sphincter EMG recording
Results
Based on the voltage response of the device an RC model was constructed to represent electrode tissue interface
and the voltage compliance for 45 mA pulses was recorded to be about 20 V. At each frequency the strength
duration curve exhibited the classic exponential decay, but when the curves of different frequencies were
compared, a repeating pattern was not observed.
Discussion
When compared with the small signal impedance of the electrodes, R in the constructed model for large signals
is considerably lower as the faradaic currents are less impeded. From the strength-duration curves, a relatively
optimal amplitude, duration, and frequency can be selected for an optimal neuromodulation.
References
[1] Craggs, M. et al, 2009. Conditional neuromodulation using trans-rectal stimulation in spinal cord injury. Neurourology &
Urodynamics, 28 (7), 836-837
[2] Graggs M., (2007). Neuromodulation device for pelvic dysfunction. World intellectual property organization WO 2007/101861 A1
Royal Academy of Engineering 28
Neural engineering - young researchers abstracts
Samantha Simons*, D.Abásolo* and J.Escudero**
As presented by Samantha Simons, University of Surrey
* Centre for Biomedical Engineering, Faculty of Engineering and Physical Sciences (J5),
University of Surrey, GU2 7XH
** Signal Processing and Multimedia Communications Research Group, Plymouth University
+44 (0) 1483 300800 ssimons@fastmail.fm
Fuzzy Entropy and Multiscale Fuzzy Entropy of the Electroencephalogram in
Alzheimer’s disease
Introduction
Non-linear analysis techniques have been increasingly reported as a way of characterising Alzheimer’s disease (AD)
from electroencephalograms (EEG), which could allow for earlier and more accurate detection. Two new methods,
Fuzzy Entropy (FuzzyEn) (Chen et al, 2007) and Multiscale Entropy (MSE) (Escudero et al, 2006) are combined in this
pilot study trialling FuzzyEn and the new combination of MSE and FuzzyEn (MSEFuzEn).
Method
EEGs were recorded from 11 patients and 11 age-matched controls (72.5±8.3 years and 72.8±6.1 years respectively,
mean±standard deviation (SD)) using the international 10-20 system with the subjects in an awake but resting
state with closed eyes. Artefact free 5 s epochs were then identified and further filtered. Both methods were
calculated with input variables n=1, 2 and 3, m=1 and 2, and r=0.1, 0.15, 0.2 and 0.25 times the SD of the time series
to normalise r, and MSEFuzEn with a coarse graining of the EEG epochs to T=12, where graining is the averaging of
successive non-overlapping windows of data of length T.
Results
FuzzyEn results showed AD patients’ EEGs had lower entropy, i.e. an increased regularity, than controls, with
statistically significant differences (p<0.01, Student’s t Test n=1, Kruskal-Wallis Test n>1) at electrodes T6, P3,
P4, O1 and O2 for the best variable combination of n=1, m=2 and r=0.15. FuzzyEn gave maximum accuracies
(proportion of correctly identified subjects) of 86.36% at electrodes P3 and O2, with reducing accuracies with
increasing n. MSEFuzEn showed controls have a more complex EEG signal except for at electrode F4 for all variable
combinations. MSEFuzEn showed greater variations of entropy along subsequent grains than previously reported
results with MSE (Escudero et al, 2006), worsening with increasing n.
Discussion
This pilot study suggests FuzzyEn and MSEFuzEn are useful methods for identifying changes in electrical activity in
the brain due to AD. However, more pathologies and more subjects must be tested to ensure the changes in brain
signals seen here are specific to AD.
References
Chen, W., Wang, Z., Xie, H. and Yu, W. (2007) Characterization of surface EMG signal based on fuzzy entropy. IEEE Transactions on
Neural Systems and Rehabilitation Engineering, 15(2), pp. 266-272
Escudero, J., Abásolo, D., Hornero, R., Espino, P. and López, M. (2006) Analysis of electroencephalograms in Alzheimer’s Disease
patients with multiscale entropy. Physiological Measurement, 27, pp. 1091-1106
Royal Academy of Engineering 29
Neural engineering - young researchers abstracts
Andrew Stewart
Neuroinformatics, Informatics Forum, University of Edinburgh
Informatics Forum, 10 Crichton street, Edinburgh EH8 9AB
0787 619 8108 andrewxstewart@gmail.com
Single-trial prediction of visual perception from EEG: analysis, classification
and ICA source dynamics
Introduction
It seems machine learning classification from EEG data can give high extracted bitrate (Muller, 2008), especially
using methods beyond traditional ERP analysis (Makeig, 2004). This has been useful for brain-computer interfaces,
but we hypothesise this can also be useful for investigating basic cognitive neuroscience, as in visual perception.
Method
We presented subjects with 50 standardised photos of everyday objects, 5 times each, while recording high quality,
70 channel EEG. We trained support vector machine classifiers on single-channels or multiple channels, using
processed EEG or component activation vectors from ICA for some approximation of source space. A ‘one versus all’
rbf SVM classifier was constructed for each object, with the timesteps 25ms-500ms (Johnson, 2005) after stimuli
presentation used for training. For robustness, we trained using 4/5 objects as positive examples, before testing
with the unseen fifth presentation of that object, iterating through each object and averaging accuracy across all
attempts.
Results
We found using a single channel vector of EEG data, we had 0.76 AUC accuracy on single trial trials, across all
subjects, correctly identifying the viewed object on around 62% of trials. This is well above chance classification
of 0.5 AUC. Each EEG channel had similar classification performance within subjects. When using ICA components
activation transforms, we found similar mean performance of 0.75 AUC accuracy, but these components were
greatly heterogeneous in their performance, with the average best component giving 0.86 AUC.
Discussion
With some real-time prediction of the current observed object from EEG well above chance, we hope to construct
cognitive neuroscience experiments to probe this. EEG data can be opaque, but in scoring the predictive power
of each of these ICA components (that give some estimate of possible sources of variance of the EEG signal
generators) we can gain a better idea of what in the EEG is relevant.
References
Johnson, J. S., & Olshausen, B. A. (2005). The earliest EEG signatures of object recognition in a cued-target task are postsensory.
Journal of Vision, 299-312. doi:10.1167/5.4.2
Makeig, S., Debener, S., Onton, J., & Delorme, A. (2004). Mining event-related brain dynamics. Trends in cognitive sciences, 8(5),
204-10. doi:10.1016/j.tics.2004.03.008
Müller, K.-R., Tangermann, M., Dornhege, G., Krauledat, M., Curio, G., & Blankertz, B. (2008). Machine learning for real-time singletrial EEG-analysis: from brain-computer interfacing to mental state monitoring.
(1), 82-90. doi:10.1016/j.jneumeth.2007.09.022
Royal Academy of Engineering 30
Neural engineering - young researchers abstracts
Balazs Szigeti
Neuroinformatics Doctoral Training Centre, University of Edinburgh
Informatics Forum, 10 Crichton street, Edinburgh EH8 9AB
The virtual C. elegans and the behavioural Turing test
Introduction
OpenWorm is an open source virtual C. elegans project that attempts to understand how behaviour emerges
through lower level biophysical processes. On top of the activity of the nervous system, it simulates a virtual
environment and the body of the worm, and the physical interaction of the two. While it is impossible to build
a model which accurately reflects C. elegans in every detail, we hope that our model will capture the essential
mechanisms that produce the behaviour.
This talk will briefly overview the challenges associated with the project, outlines OpenWorm’s strategy to
overcome these difficulties, and attempts to define the first few steps on the long road towards a model that can
pass the behavioural Turing test.
Discussion
C. elegans is one of the most studied multi cellular organisms: its genome has been sequenced, studies of its
programmed cell death and regulation of organ development resulted in a Nobel prize, and it is currently the only
organism that has had its nervous system’s wiring diagram mapped [1].
Inspired by the fact that is has been so well studied, there have been multiple attempts to create a virtual C. elegans
during the last decades. These projects achieved limited success, but nonetheless greatly helped the scientific
community to appreciate the depth of the problem. It is a humbling lesson in biological complexity that despite
decades of extensive research we are still far away from understanding the structure-function relationships
between C. elegans’s nervous system and its behaviour.
Norbert Wiener, the founder of cybernetics, famously said that ‘the best material model of a cat is another, or
preferably the same, cat’. This argument is frequently used to question the rationale of creating semi-holistic
biological models. The virtual C. elegans is not a tool to study the worm itself - for that, the worm is really the ideal
system - but is rather a challenge to drive forward biology and computer science. On the way, many issues could be
clarified, such as about the relationship between the model and the organism, the benefit of connectome datasets
in neuroscience, and the redundancy of biochemical processes in computer simulations of behaviour etc..
The importance of chemical signalling networks in behaviour will be discussed first, followed by the generalisation
of Turing test for biological systems [2]. In the second half of the talk the current status of the OpenWorm model and
future plans will be addressed.
References
[1] Varshney LR, Chen BL, Paniagua E, Hall DH, Chklovskii DB (2011): Structural Properties of the Caenorhabditis elegans
Neuronal Network. PLoS Comput. Biology 7(2):e1001066.doi:10.1371/journal.pcbi.1001066
[2] D. Harel: A Turing-like test for biological modelling, Nature Biotechnology 23(4):495-496, 2006
Royal Academy of Engineering 31
Neural engineering - young researchers abstracts
Ian Williams
Centre for Bio-Inspired Technology, Imperial College London
i.williams10@imperial.ac.uk
Funded by EPSRC
Proprioceptive feedback for prosthetic arm users
Introduction
In recent years, there have been great developments in the capabilities of powered prosthetic limbs. However, a
major barrier to amputees accepting and using a prosthetic limb is the lack of sensation they have in the limb and
the challenges this creates for them in controlling it. In other words they lack proprioception – the body’s ability to
sense where the various parts of our body are in space and the forces our muscles are exerting.
Method
In the human body, two of the key neural receptors for proprioception are muscle spindles - which provide
information about muscle length and as such limb position and motion - and Golgi Tendon Organs (GTOs) which
provide information about muscle strain.
Our work will initially focus on the flexion and extension of an elbow joint of a representative prosthetic arm.
This elbow joint will be fitted with angle and torsional strain sensors which will be used to estimate muscle lengths
and strains. This information will then be fed into neural models of the spindles and GTOs to produce neural firing
patterns analogous to those found in the human body. Finally these neural firing patterns will be passed to a neural
stimulator which will stimulate neurons in the peripheral nervous system accordingly.
Results
The initial phase of my research has focused on determining the various requirements for a proprioceptive
prosthesis and developing a low power, safe and fully integrated neural stimulator suitable for this application (see
reference). Currently I am working on system level and experimental design and developing non-real time, PC based
implementations of the processing block.
Discussion
The aim of my research is to develop and validate the fundamental blocks for a proprioceptive prosthesis.
Thereby investigating the various tradeoffs and identifying the remaining issues that need to be addressed to
create a device that we believe will enable prosthetic arm users to control their limb more naturally, more accurately
and with less cognitive effort.
References
I Williams, TG Constandinou, “An Energy-Efficient, Dynamic Voltage Scaling Neural Stimulator for a Proprioceptive Prosthesis”,
International Symposium on Circuits and Systems (ISCAS), 2012.
Royal Academy of Engineering 32
Young researchers futures meeting
Neural engineering
19-21 September 2012, University of Warwick
Feedback form
Full name (you can be anonymous):
University/research centre:
Please rate the following using the scale:
1
poor
2
below expectation
3
acceptable
4
good
5
excellent
The registration/admin process
Venue/accommodation /catering
Quality of the speakers/mentors
Networking opportunities
Relevance/benefit to your research
If you have rated anything as poor or below expectations, please explain why:
What did you find most interesting/useful?
What did you find least interesting/useful?
What activities could the Panel for Biomedical Engineering organise to help promote research and inspire the next
generation of neural engineers? (Note: we cannot provide research funding)
Other comments:
Please leave the feedback form with Professor Christopher James before you leave.
Alternatively email the completed form to c.m.shepherd@warwick.ac.uk or post to:
Carla Shepherd, Institute of Digital Healthcare, International Digital Laboratory, University of Warwick, CV4 7AL.
Royal Academy of Engineering
As the UK’s national academy for engineering, we bring together the most
successful and talented engineers from across the engineering sectors for a shared
purpose: to advance and promote excellence in engineering.
We provide analysis and policy support to promote the UK’s role as a great place
from which to do business. We take a lead on engineering education and we invest
in the UK’s world class research base to underpin innovation. We work to improve
public awareness and understanding of engineering.
We are a national academy with a global outlook and use our international
partnerships to ensure that the UK benefits from international networks,
expertise and investment.
The Academy’s work programmes are driven by four strategic challenges, each of
which provides a key contribution to a strong and vibrant engineering sector and
to the health and wealth of society:
Drive faster and more balanced economic growth
Foster better education and skills
Lead the profession
Promote engineering at the heart of society
Panel for Biomedical Engineering
Engineering is at the heart of society, underpinning and continually improving the
quality of our lives. The Royal Academy of Engineering brings together the country’s
most eminent engineers from all disciplines to promote excellence and support the
engineering performance of the UK.
Biomedical engineering creates new medical technologies and systems that can
greatly improve patient care and quality of life. The Panel for Biomedical Engineering
is the Academy’s forum for this increasingly important area of engineering in which
the UK is taking a lead.
Royal Academy of Engineering
3 Carlton House Terrace, London SW1Y 5DG
Tel: +44 (0)20 7766 0600
www.raeng.org.uk
Registered charity number 293074