0:00:15okay so um the next speaker is
0:00:18and robert rubber smith
0:00:19oh she will be presenting
0:00:21a a a a method
0:00:23for robust minimum variance beamformer and its application
0:00:28to um and E G and
0:00:30a little potential
0:00:54i can start
0:00:56the um that top like a to see it
0:01:00okay a i it
0:01:01that's that some work which we do doing in oxford
0:01:05between a of us
0:01:07so i i mean miss amy
0:01:09is the map rotation any he's done most of the work on the details the beamforming
0:01:14so i look a fast for that
0:01:16we also working with um the pulse like a tree
0:01:20um i an engineering
0:01:22that's made
0:01:23and speaker i Z is is one of the D sections
0:01:26work in they uh that if possible source just or not
0:01:29in the brain stimulation
0:01:32this is a the based in a sense
0:01:35we've got a clinical problem
0:01:36um which is to do with the brain stimulation
0:01:40and we are trying to image it's using a make and and stuff a
0:01:44so i'll start with a base
0:01:46discussion about um
0:01:47and then all want to the beamforming and then out of use some results
0:01:53a just a fish
0:01:59that that stimulation is technology
0:02:02web i
0:02:03oh they plan electrodes into the brain
0:02:07so that the here
0:02:09you can see a
0:02:12that that coming in a top
0:02:14coming coming down
0:02:17it's used
0:02:18white and not your the sources
0:02:20in particular we you it's not with problems
0:02:24for the most common
0:02:25uh those that eating with M are that is press it's channel
0:02:29and eating with a a a a a motion problems slide stand here like can easy easier
0:02:35case not can be talking about that is pain
0:02:39and we say that a uh is that you've a person
0:02:42well pain is for the the perception of pain is all sixteen is subjective
0:02:47some people see it could very well with pain or not find a right
0:02:52um it's also use some of the
0:02:54um areas such as much as its is not in your
0:02:59so no one yeah
0:03:00stands for it works
0:03:02what you do is you plot the election in the brain
0:03:05and will be here brain stimulation the of signals actually coming out of N be very small
0:03:10the electrodes implant
0:03:13they then put on a five vote
0:03:15sine wave normally five votes like very bits
0:03:18at a frequency to about fifty hz and the functions that a
0:03:23um and then a a a go through period of this a a a lecture it's by search
0:03:28that's then a period where the external eyes
0:03:32well you might access to the stimulates to
0:03:34and S the signals coming off the implanted electrodes the local field potentials
0:03:38and during that time they tried rate
0:03:40so they one the site that's signal of the patient
0:03:43and then off to that they employ a battery
0:03:45and everything missing the was down inside everything becomes internal
0:03:50the what we do is very trying to get a of that text last period
0:03:54so we can actually the feel potential of the electrodes
0:03:57and use the
0:03:58to try to improve a
0:04:02so mad ninety two and stuff lot of a
0:04:04a thirteen new technology
0:04:07picking up
0:04:07a magnetic fields of you were have to um
0:04:12so that that's
0:04:13about three more three um eight or the magnitude smaller than the spec
0:04:19so essentially put a number of my comment isn't but you might as rather here
0:04:24a she
0:04:25you can see what here
0:04:28and for these very small signals
0:04:30then we tried to um we can sort the sources within the brain
0:04:35and a point of that may is to so
0:04:37uh one is to improve
0:04:38the surgery so they it that what in four
0:04:41and the other one is to try to understand more about the you it's
0:04:48so that's what are difficult to use
0:04:50with D B S
0:04:53one is way looking at that very small signals
0:04:56we expect most of the
0:04:58excitation of the brains to come from the a region around the a lecture
0:05:03well fortunately
0:05:04that's quite good separation between the frequency at which are the electrodes
0:05:08and the frequency that which principle
0:05:11so it is possible to try to access that's small signals want to me like sing
0:05:16so we looking at about say fifty to a hundred thirty hz stimulation
0:05:20and the low cell
0:05:21um a a range between about five and fifteen a
0:05:24for the range from
0:05:28yeah that that quite big it's you
0:05:30is when you bring the wires out
0:05:32they are have to do a lot and skull
0:05:35uh the wires a magnetic you got cost of wires are yeah the whole of the skull than not done
0:05:41but those double for distortion
0:05:44a that region in terms of your source
0:05:52so that think in this paper a is where
0:05:55um got two things it using a placement of an now now that a whole
0:06:00a a well on that show how to improves
0:06:02the recovery of the of the
0:06:04spatial source
0:06:07what we think to take that in this paper
0:06:10it's it's used in the idea that want all simulation is on
0:06:14then it's
0:06:15you would expect most of your signal
0:06:17so yeah the stimulation can see
0:06:19be coming back from a region
0:06:21and which stimulating
0:06:23and hence by splicing the cross correlation
0:06:26between the signal you putting in
0:06:28and the signal you're thing
0:06:30at my comment is
0:06:32then any fact you can improve your beamformer
0:06:35especially one region of interest
0:06:37which is found the region whether you like to tune
0:06:41so the idea here is way looking at stimulation on and trying to get the best stuff beam as we
0:06:47and then but using
0:06:48time and stimulation is all
0:06:50to try to verify that
0:06:53so what i do i'm elements is they stimulate the time
0:07:00people don't seem to get a
0:07:01uh used to that's
0:07:02T where king
0:07:04but there are it's use a at that time
0:07:07so that just cost about ten thousand pounds
0:07:10and they lost from a couple of years that's what made thing not to take infection
0:07:15the um
0:07:16put in you new battery
0:07:18yeah that is you
0:07:19which is
0:07:20coming more to like that does seem some evidence that stimulation at
0:07:23that some of the of the range emotional areas
0:07:27and that's some showing that actually and not to stimulate send march
0:07:31because of things that a
0:07:34so a of my and fast which should welcome foundations support
0:07:37is trying to actually um
0:07:39find an adaptive method where we actually you know way to simulate when
0:07:45once tried in former to you
0:07:47very quickly um Z say that's very high meets work and i don't think that yeah
0:07:52that you
0:07:53and then began to demonstrate using training data
0:07:56from a a a a a patient
0:07:58with P B S i mean have them
0:08:09so a techniques
0:08:12yeah shows um is look something like
0:08:15so that that's uh how a bit of each
0:08:18and to
0:08:20a few centimetres long
0:08:22they have a number or electrodes or may have about four electrodes on where you can stimulate
0:08:28these are the to that the field of getting around each one is in fact a let's
0:08:35so we can to assume the N F P data comes from a small cool volume
0:08:39a bound the implant electrode we know that where that is from looking at um
0:08:43ct a oh i want to implement
0:08:46you can use
0:08:48um we can we that now
0:08:49a a a a a whole of skull
0:08:52and then make a to use a robust minimum variance beamformer
0:08:55um what's the to work by element set out in two thousand and five
0:08:59but the obvious didn't use the particular aspects which relevant to um make imaging
0:09:07yes or forward model
0:09:11so the first time here
0:09:15so so white state wide C R be of measurements
0:09:18at skull
0:09:20um S
0:09:21is the stimulation
0:09:23so that's from the electrode
0:09:25and a and is the forward model leave but a vector
0:09:29then we have a second term
0:09:31which includes the
0:09:33uh oh S to try to now but i
0:09:36that's a N
0:09:37as T
0:09:40and we assuming a um but the N F P data is going to a small volume only got here
0:09:46i can the location of the whole is known
0:09:48so we can deal with that
0:09:50and then got some noise
0:09:52there's also in fact times day
0:09:54going from be um lot because the instrumentation
0:09:57between the
0:09:59um excitation signal
0:10:00and am signals with a
0:10:05so just think they are a problem formulation
0:10:07is to optimize are beam
0:10:12um so we've got a why we want to estimate S
0:10:17use optimisation on it
0:10:20or or first time here is the difference between that's essentially
0:10:23thank expectations
0:10:25um with a value you out for a penalty factor
0:10:29um which there is we be and one
0:10:31so you put note you want you take in and that is
0:10:34of of the product of a correlation
0:10:36between the source
0:10:38i'm the measurements
0:10:40and then subject to all source vector
0:10:44so that's well
0:10:46and it to it's all
0:10:47so we know we got source that
0:10:49is the only condition
0:10:51okay and then we find that
0:10:57so that if we take the um the variances of the and take the expectations
0:11:02then a first term does becomes
0:11:05this one down here
0:11:07so a why S yeah is the first good that is
0:11:10between the source
0:11:12the excitation we put on and the measurement
0:11:17it's a a a a solution essentially
0:11:20well we're trying to maximise or or correlations so
0:11:23something to minimizing constraints such that constraints
0:11:26you can find exact
0:11:27use the W
0:11:29this for increments as work that E
0:11:31which comes to stop one
0:11:33yeah yes one of the parameters which says what which is the described in that
0:11:37where you putting on the excitation
0:11:40now to me a of once month i
0:11:44um if we substitute are right if W back into the a constraint
0:11:48and in fact
0:11:49we find you
0:11:51this is but
0:11:54we then that no as that's
0:11:56using the
0:11:57normalization low
0:12:00so be yeah so i put in this
0:12:03um we
0:12:04oh of the um diarization is with that i've the factor
0:12:08are a it's a half sheet
0:12:11a a heart sounds
0:12:13so that's the vector which later
0:12:15the region of what we know the excitation is
0:12:18to the
0:12:19um close correlation between the measurements and the excitation
0:12:23we can do that i was that get a secular equation
0:12:29which late
0:12:30a a a a me we found this relates number of than our on unknown
0:12:35to the eigenvalues yeah i
0:12:38of all normalized
0:12:41or a of how cute
0:12:46so the details of that are in the paper
0:12:48um and the was in the paper and so i am of a fast of it that that's
0:12:52time is and results
0:12:53um and i have got some papers is here no once
0:12:56but a big advantage the way that actually be normalized this
0:13:00so we've normalized here
0:13:03talk about V before
0:13:05so E and
0:13:12with the to some of those covariance matrices
0:13:14and also the parameters of the of is what we put X
0:13:21is actually uses as a nice solution which then amenable able to an effective
0:13:25solution to have find
0:13:27um and uh
0:13:28or as some previous work has a a few a given bounds for them the
0:13:32this kind of normalization as she allows as to optimize and you know
0:13:39a out a which allows us to do that
0:13:41which are not been to go to i think
0:13:44that that used to some themselves
0:13:47so we've done this in simulation one of the problems of course as we all know made doing things well
0:13:52mel by medically
0:13:53um it is very hard to try to validate
0:13:57so we done validation using a simple simulation of a spherical head model
0:14:02put in the deep source
0:14:04simulates the excitation station
0:14:06we put see in the source of the but of interference
0:14:09um and with that's noise
0:14:12and to sure that we can to look up on and all conditions
0:14:16we've allowed to all source
0:14:18um which would only be
0:14:20i dominate dominated
0:14:21by the stimulation we on
0:14:23or we should be seen was once the brain when off
0:14:26a different frequencies you sign "'cause" on way
0:14:30so yeah that was with the start of wiener filter a
0:14:33and we've compared them with the filter so in both whole now without correlation
0:14:40so yeah it is just the interference and their noise
0:14:44i i mean is
0:14:46um because it's that we've
0:14:48simulate interference long that thing else be be estimated it
0:14:52um we can see the wiener filter
0:14:54is that you not
0:14:55um a well
0:14:58um um
0:14:59as well
0:15:01improves things the the S I
0:15:03but that but
0:15:04using the be
0:15:09so if that the S noise as well as a fair comparison
0:15:13that we can see a method that she does give us advanced just
0:15:16oh the the other two men said putting in this post correlation so um
0:15:19um between of the source
0:15:21and the measurements since we know sources is doing that frequency as she does hell
0:15:28a technical data is most interesting
0:15:30um of the forty or or one with a body pain wanting pain had separate yes
0:15:36um in but in the pack
0:15:38which is close down to them
0:15:41oh and fifty hz
0:15:43so the it and sell them on magnetic
0:15:46so they are a problem
0:15:47for the mac
0:15:50but in fact
0:15:52say whether whether wow clusters then you can still
0:15:57do you have an all i can just and that's
0:15:59before four we assessed the lectures
0:16:01for sequences like to
0:16:03um so i are you can use for but not often
0:16:06you got inside you
0:16:08and that just can't of is that um
0:16:11using the he was first right
0:16:13this my
0:16:17the things we do
0:16:19first of all
0:16:23recall yeah that P data in all conditions
0:16:27so that's that she recording straight of the electrodes
0:16:29the that are too long used a beamformer
0:16:33and then we were what happened all
0:16:35so we use the beam forming one number you know
0:16:39we then be constructed using a P point how much is what we would expect to see from
0:16:45the yeah it is maybe a filter
0:16:48yeah it is
0:16:49using a a a a as well
0:16:51a use of the um i three
0:16:53not very much
0:16:54and here it is
0:16:56using you a
0:16:57and you be
0:16:59so you can see you press that i think that that i
0:17:01um i think this is a much better
0:17:03the U B
0:17:08the a point of this
0:17:09try to find out
0:17:10a in the brain
0:17:12so this is rate just the fit it interest as suppose
0:17:15and this is a get the difference
0:17:17between the
0:17:18stimulation of
0:17:20so this woman that
0:17:21a once it was turned off
0:17:23and you didn't
0:17:23on on
0:17:25um not an entirely goods
0:17:27um experiment because she stopped yes you knows this awful
0:17:32you a you do get some differences between the of a young condition
0:17:35and the way in which are response
0:17:38this a you looking at part should known to be sensitive to pain
0:17:41um um rate standard high
0:17:44a i'm the A C C is the and tear single call
0:17:47yeah right
0:17:49and the three and i is the um of the email
0:17:54so this technique can see that point
0:17:57spatial layers grain
0:17:58um where we getting sports
0:18:00to be um nipple stimulation
0:18:04so we we've also a of using correlations that all closely and identify more closely spatial regions in right
0:18:11um i i a and the results if five using elliptical volume of the yeah that's P
0:18:16to improve things of using circular volume
0:18:18still provide
0:18:20and we've shown some results
0:18:23improvements using simulation
0:18:24i mean and you
0:18:26and you
0:18:49as much
0:18:52or a more you are still going on
0:19:01or you are you
0:19:03two more
0:19:14or like to call
0:19:20yes may just the problem
0:19:21say that's that
0:19:22for my my kinetic model of the brand such a simple
0:19:26"'cause" you can assume that is a of a two D is
0:19:29is one is what
0:19:33oh well as for a um but it's not a context could you what a much more complex model right
0:19:38so my a standard are not experts what they don't E G my understanding is make use gives you what
0:19:45that is to discuss that with you for are actually that if anyone and any information on that
0:19:49that's a my hands and you i mean it is a recent technology is come about lost five to ten
0:19:54well a you also
0:19:55to machine
0:19:56and not let me
0:20:06what should you signals
0:20:13what was is
0:20:15are you sure from your speech and so
0:20:24for a the S T N was a packet is next to it
0:20:30but pain yeah
0:20:31the use S the end from it
0:20:33oh so you're right
0:20:37trying to be your with peace
0:20:40for four
0:20:42what what you saying
0:20:43for the final are we measure yeah that peaks as X to i and then we tried to estimate as
0:20:48yes using the mac
0:20:50in a compared
0:20:55for point
0:21:00so what is
0:21:08the marks is
0:21:10i five to
0:21:11i i mean that sense that to the signal
0:21:14we receiving see in the of
0:21:15or you
0:21:16a frequency
0:21:17a point where we stephen it is that you want place that
0:21:21is quite hard to know how you try to optimize
0:21:24world i mean would really strong
0:21:27so i
0:21:29one Q one
0:21:38which still
0:21:38what we are we are doing a spectral analysis as well so we looking
0:21:43so a beamformer is that you
0:21:46i i think that if
0:21:49so we all by some that to look at
0:21:51maybe to right
0:21:52the B two
0:21:53of what's it's yeah was either
0:21:58and yeah the
0:22:06source you right hmmm part
0:22:13yeah a just that even that's one
0:22:15that's what if you they get
0:22:18sure one
0:22:20uh_huh four
0:22:24we would
0:22:28it's a that's not in a sense
0:22:30that are not quite so what you're looking for
0:22:32seven seven sense i mean it is that's what we try to estimate we getting and then measure we get
0:22:36"'em" and i suppose way we can possibly indicate
0:22:40one thing to get a good results or
0:22:42but it's good question and very hard to know how about it
0:22:49you try to the um
0:22:51so i mean
0:22:55okay well thank you everyone
0:22:57uh i think you again maybe