<TITLE: Genetics and Pathology of Oscillatory Brain Activity
ACADEMIC DOMAIN: medicine
DISCIPLINE: neurology
EVENT TYPE: lecture
FILE ID: ULEC180
NOTES: lecture interspersed with some discussion, continuation of and continued in ULECD110

RECORDING DURATION: 47 min 49 sec

RECORDING DATE: 29.3.2007

NUMBER OF PARTICIPANTS: 13

NUMBER OF SPEAKERS: 3

S1: NATIVE-SPEAKER STATUS: Finnish; ACADEMIC ROLE: junior staff; GENDER: male; AGE: 31-50

S2: NATIVE-SPEAKER STATUS: Danish; ACADEMIC ROLE: junior staff; GENDER: male; AGE: 31-50

S3: NATIVE-SPEAKER STATUS: Finnish; ACADEMIC ROLE: research student; GENDER: male; AGE: 24-30

SS: several simultaneous speakers>


<S2> thank you <NAME S1> erm i promised er <NAME S1> to make a couple of comments about so i i did my PhD in in helsinki and i moved to amsterdam to centre for neurogenomics and cognitive research erm which is a fairly new interdisciplinary research centre where we try to or we have here the ambition to at least contribute to the development of er integrative neuroscience rooted in molecular biology and genetics and extending all the way up to behaviour cognition in in er healthy subjects and er patients , and the way that we do that at CNCR is to motivate a closer collaboration between neurobiologists clinicians psychologists and geneticists erm in making them generate er joint research programmes and er , so as all of you know erm if you want to understand the workings of the brain you can do that at many different levels you can focus on genes that whose expression is relevant to brain function proteins synapses nerve cells networks systems neuroscience behaviour both in animals and humans but typically you divide er neuroscience research into di- distinct departments that that specialise in each of these different levels so (how) and often researchers on one level er have difficulties communicating or even finding a a shared language with researchers in another er department , so at CNCR we we de- or they decided to er to take the gene as something we have in common because gene is a , er genes do er make a contribution to both the structure and the activity that you can er investigate at all these different levels and to facilitate erm that everybody at these different in these different departments erm have a genetic perspective to their research erm both the behavioural and in vitro er neurophysiologists that used to work in rats are now moving to mi- towards getting their research up and going in mouse lines because that was used more by the molecular biologists erm we're trying to instead of taking care of our own funding and then meet and not communicate er money's always a good stimulus for breeding conversation er so there are more proposals submitted across departments and erm , and then by focusing both animal and human research on the same behaviour or the same disease that also facilitates er communicating across departments and maybe especially the latter is typically a- seen as a big jump so if if you work in a in a human systems neuroscience lab er there's a great hesitation for those researchers to get involved with animal research and vice versa but because of our focus on the genes and the presence of er the netherlands twin register we have those of us who do human er systems neuroscience have the opportunity to contribute erm to our u- understanding of of genetic basis of of er of systems level brain structure and brain function because when you compare groups of monozygotic and dizygotic twins you can estimate er the contribution of genes to the variants of whatever trait you are looking at be it the thickness of the cortex or the amplitude of an oscillation or frequency et cetera , and because we have so many twins and they had a tradition for following these twins every second year throughout their life it also provides a very valuable er subject database for recruitment to er er to basic research and erm one recent success of this is when the US had a very ambitious erm project for finding genes er involved in depression they needed er not only depressed patients but also control subjects and there in previous research it's often so that you take in random people who just say no i've never been depressed but you enter all the people who have been 90 per cent depressed or who have not been depressed but their brother has been depressed which creates a lot of noise and because of them having followed these twins at two year intervals including their siblings erm for many years they were able to extract (to offer) this US consortium to come up with 2,000 healthy subjects that for sure never had been sub- er depressed and their family members had not been depressed either and that way we got funding for genotype for a thousand subjects , another strength i think it is that both people who are working at the molecular level neural and network level mouse behaviour and human behaviour erm are moving from small scale studies where you have ten mice or ten human beings to er medium (through good) level where you er scan the behaviour of a thousand mice or a thousand people er or a thousand brain slices . so in the department of experimental neurophysiology where i'm currently sitting in the neuroinformatics group what we have done to scale up the the speed of of acquiring data is to <COUGH> set up erm a four time er 16 er channel multielectrode array where we can measure er oscillating from brain slices and that allows us to actually measure er in vitro oscillations from four slices and the changes (we have) six mice per day so we can get more than 20 slices erm measured a day and er we have then developed software for tissue typing where anatomically the different channels are (sitting) so that we can enter the data into a database and and er fairly automatically er characterise the oscillations . this was just a short introduction to some of the activities er taking place at CNCR er integrating activities across different levels in neuroscience is not the only problem of interest it's also interesting that activity in the brain needs to integrate itself at different levels er many of you have probably heard that that the neuronal network oscillations have been proposed to play an important role in integrating activity at different levels and at least if you want to do integrative neuroscience oscillations have this nice property that you can both measure er in vivo in humans and see if development correlates with cognition and performance but it can also be studied at lower levels intracranially you can measure local field potentials you can me- er measure single unit activity and see oscillatings in in their firing patterns and the exact shape of the oscillations depend on the cell types the microcircuitry of the cortex er the synapse currents you can er genetically or pharmacologically modify synapse currents and see investigate the influence on oscillations so by focusing on oscillations we also in many departments have a shared interest , and that is then where my work starts i'm also interested in oscillations and that i have been for many years and now i'm i'm adding a a genetic perspective to my research , but how do we how have we previously analysed oscillations everybody knows that the human er EEG oscillation is has kno- been known er since the 1920s that if you record a human EEG oscillating it looks like this but until recently the way that you would actually quantify er this signal is to make a power spectrum you in a sense make a description of this signal that looks like this and then you quantify what is the frequency and what is the amplitude the question is okay there's probably the possibility of a bit more information in this signal than in this signal so i have been interested in what possible information could be embedded and how these oscillations are modulated in amplitude should we see the modulating as just stochastic environmental influences we know that sensory stimuli for sure can im- have an impact on what happens to an on-going oscillating or does it happen intrinsically because of biological mechanisms and what does this complex temporal modulating how could that relate to cognition and can we make computational models of it that's what i'm going to talk about in the rest of my talk i should say that i think we are so few people that if there's something not clear then just raise your hand and and ask me to explain better er 'cause i think many of you are not in systems neuroscience many of you are maybe not even studying oscillations <P:13> well studying the frequency and er average amplitude of oscillations can be done with (xx) are a little bit more sophisticated and what has become popular with the past ten 15 years is to investigate the phase of an oscillation measured in two different regions is (xx) that you can do by coherence or a phase locking analysis for instance so you make a linear correlation analysis between the phase here and here and that is has been seen as as as interesting because it could be that er phase correlations between oscillations in distinct brain areas erm is a way to deduce whether these different areas are involved in the same processing tasks so it's in a sense a way of deducing whether there's parallel processing going on , what i'm my research is digging into and which not so many people have taken serious so far is whether we also by studying correlation over time can deduce serial processing so in a same way that different brain areas are involved in a certain task it's also clear to me intuitively that at a given point of time what the brain is doing is related to what the brain was doing two five ten and twenty seconds ago and it is not implausible that correlations in the modulation of beta oscillations could reflect this , i will come back to that later erm , first i want to show how we could possibly study this so what wh- what i did when i was working in in helsinki is to record er long sequences of spontaneous brain activity on humans with an M-E-G device and if you then take the amplitude and (xx) oscillation (out) and you view 150 seconds then you get this complex pattern , and at that point of time we did now know if there was any information embedded in this structure and we then tested three different correlation analyses so we're not correlating two different signals we are correlating the signal with itself at different time (laps) to see is it correlated or is it uncorrelated if it is correlated we could have some information if it is uncorrelated statistically it is not so likely to have information and what we see is many of you might be m- mostly familiar with the outer correlation function what you see when you do an outer correlation analysis of this signal is that the correlation decays rather slowly towards zero 'cause this is a logarithmic scale so if the de-correlations remain significant up to 200 seconds this is quite a lot because in time scales of the oscillation that corresponds to more than a thousand oscillating cycles somehow (xx) each other a little bit , you get a similar result if you do a power spectrum analysis on this signal or if you do e- employ a an algorithm called DFA i'm not going to explain this algorithm in detail even though that's the one i use in the remainder of my talk it is an algorithm that is it does not have as strict assumptions as the outer correlation analysis and still allow you to quantify to what extent or how quickly er these correlations decay towards zero , and it works so that if you do a DFA analysis on random data then you get a so-called DFA exponent of 0.5 if it's random and between er 0.5 and one if there are er significant correlations <P:10> i'm still gonna ask once @again@ @@ is is this er 'cause this is where i usually lose my audience @@ is this completely clear that that you can have an oscillation that changes amplitude in a fairly complex manner over time and you want to to ask a question are the fluctuations here somehow dependent on the fluctuations here at a certain layer to do an outer correlation analysis on that and then you find that indeed the fluctuations here are not independent of here the correlations are weaker than if you look at a shorter time scale so it's natural that if you go to longer and longer (laps) er the correlations wi- wi- will be weaker but it's important that they do not become zero very quickly if you (xx) random data and do the same analysis you see that after two seconds correlations are at zero , so this temporal structure has the potential to to contain some information but again it could be environmental stimuli that come to the subject in an structured manner and modulate the rhythm so that it is not pure biology that we w- that we see reflected in the strength of these correlations so that is what has motivated us to study this in a large number of twin subjects we have measured eyes-closed rest EEG spontaneous brain activity in almost 400 twins and then we can er make a heritability analysis that tells us to what extent the proportion of phenotypic variance so phenotypic variance is the the variance in long-range temporal correlations from one subject to another to what extent that variability can variability can be attributed to genetic variance and that you can do when you have twins because there's less genetic variance in the monozygotic twins than in the dizygotic twins <P:08> i have to mention there's one , difficulty or time consuming (so far) erm aspect of doing this long-range temporal correlation analysis and that's the artefact you get you have an eye blink er it creates so much temporal structure in your data that you will not only see the temporal structure of the oscillations but also of whatever shoulder movements eye blinks or maybe even heartbeat you sometimes you can have that the heartbeat shows up also in the EEG recording , erm but fortunately we have algorithms now that can blindly separate erm multichannel data into distinct source topographies and then the topographies can be identified with certain artefacts so that a topography like this would typically be an eye artefact whereas a topography of this would typically be er an alpha oscillation from the parietal region and then you can project those topographies out that corresponds to your artefacts and then this noisy data actually looks this clean so you can fairly selectively get rid of the artefacts , and then you can proceed with your analysis you (xx) the oscillation you take the amplitude (xx) this is on a fairly short time scale ten seconds on a longer time scale it's one hundred seconds you see the complex fluctuations in two dizygotic twins and already if you shuffle this signal in short windows then you get this , so you can already see that if you ha- randomly shuffle the same data the t- temporal structure is quite different that is then reflected in the DFA er the exponent becomes 0.5 for the shuffled data whereas the original data had an exponent close to one but this dizygotic twin brother is in between the fairly strong correlations seen up here and there are no temporal correlations down here we want to know what is actually causing this difference <P:06> i should say that that the amplitude of the power (of alpha) oscillating cell has been known for some time to be re- highly heritable so this is a knox scale so actually that has a in in a in a normal population of subjects the power can vary (a factor) of more than 20 that's completely normal er but er but it varies with strong genetic influence so monozygotic twins you know the alpha power of one twin er you can predict with very high accuracy what the power will be in his twin brother whereas if you look at the same scatter plot this is again the alpha power of one dizygotic twin against his twin brother because they only share half of the genes there's a much greater spread here than for the monozygotic twins , so when you compute the heritability for alpha power it's actually as much as 85 per cent that is er determined by genetic influences . so the DFA exponent which is indexing the long-range temporal correlations we see here a similar effect there's a much stronger correlation er between monozygotic twins than dizygotic twins and if you formally er model to what extent this variability is determined by genes you arrive at estimates between 40 and 60 per cent . now i think it's healthy to b- be a little bit sceptical when someone says er i i have an algorithm that somehow measures correlations in in this mess er and it's genetically determined erm that's you can see in in my scatter plot because otherwise there wouldn't be a correlation between the monozygotic twins <COUGH> it's always nice if you can go back to your original signal and actually with your eyes see what your algorithms are and i think that's quite striking so what you have up here are two monozygotic twins you see they have very similar alpha amplitude and the temporal structure is very different they have small DFA exponents so there's not a lot of erm temporal structure in their oscillations if you go to those monozygotic twins that have er a lot of temporal structure a lot of temporal correlations then you see that yeah if one subject has very long-lasting modulations then his twin brother has it as well down here it's also clear that these monozygotic twins they look very similar to each other compared to this twin for instance with any of the other subjects <P:06> mhm . when you then study such a trait there's always er a danger that you are just inventing an algorithm that somehow measures an already known er aspect of oscillations er for instance if you study coherence between two channels that are close to each other and just for for the reason of ion conduction then coherence will be highly heritable because coherence is not independent of amplitude erm but the DFA exponent doesn't seem to suffer from this problem because if you correlate er the power of the alpha oscillations and their temporal correlations you really get zero correlation which is highly non-significant with all electrodes even though and we have quite good statistics for saying that so to conclude erm , the long-range temporal correlations of the DFA exponent that we use to index there erm this has a fairly high heritability and er it's independent of amplitude so it sort of has the potential to be an additional trait for characterising neuronal network oscillations erm and it is yeah because it is genetically highly influenced by genes er long-range temporal correlation analysis can er act as an endophenotype to bridge the gap between er oscillations as a covariant (xx) in disease er and that is then what i want to move on to i- in the next study where we show how these correlations can be related to disease are there any questions </S2>
<S3> i just <S2> [yeah] </S2> [want to] ask about the er effects of morphology of the scull since twins have the same shape of the scull and if you are comparing them is it more easier than to compare two different types of sculls and i mean where do you measure these oscillations could it be , i mean what is the shape if you just like a few inches away are measuring and , are there , how carefully you need to put into the same position these er electrodes </S3>
<S2> erm . it is true that well there are two things to say erm for the heritability estimates of of alpha power for instance this is definitely something that er could contribute to the high heritability so it's because the the the amplitude of the signal is also influenced by the thickness of the scalp scull for instance then er if that yeah so if the scull thickness is heritable that will spill over in the alpha amplitude also being heritable erm and i was thinking about DFA this has it is not er sensitive to the actual value of the signal but only how it is modulated <S3> oh </S3> erm , this is also why DFA is suited for a comparison between the people that are er studying oscillations for instance in vitro and in and in vivo er or even the EEG and M-E-G er because it has to do with how the signal is modulated over time er it is not sensitive to what the unit is . does that sort of answer your <S3> [yes] </S3> [question] yeah <P:10> okay so now we have sort of established DFA as as as a quantitive trait erm meaningful bilateral trait to use for characterising oscillations i would like to show a couple of results from the john lisman's lab er for how how er oscillations in different brain regions and different frequency patterns er are changing amplitude as a function of er memory and and attention (xx) so if we have a a sternberg working memory task where you are sequentially presenting the subject with different items that they have to remember and then there's a delay period so-called retention interval before a probe item is presented and then they have to decide whether this probe was present in these items and of course they can only respond correctly if they have actually er held these items in working memory so what you see during this interval of encoding and keeping items in memory is the different oscillations so that period is indicated by a black bar down here so in frontal region you see that the beta band is elevated in this period some frequency band in-between beta and and alpha is activated and down here er in the temporal lobe there is quite (sub-sustained) activation of the alpha band and very importantly in in i- if you look at temporal oscillations in the parietal lobe and you vary the duration of this encoding and delay period you see a parametric increase in the duration of how long are these oscillation er how long do they stay elevated in amplitude . these data are from intra-cranial recordings so there's such a high signal-to-noise ratio that you can really in the raw signals see that so this black bar that's one they have to keep items in working memory you really see how the amplitude goes up during those periods and down when they present the probe <P:06> mhm-hm , so this has led us to hypothesise that mnemonic and attentional processes require sustained oscillations to (bind) information across tens of seconds ma- many people have suggested that the phase of the oscillation for instance is important for maintaining a a neuronal representation but there is quite little emphasis on the fact that that phase of the oscillation is not there if the oscillation is not there you don't (get to) find the phase of an oscillation if the amplitude is not there so if you need to keep an item active in working memory for many seconds you probably need the oscillation to be elevated in amplitude also for several seconds <P:06> mhm-hm . but then going back and thinking of our data where we often look at spontaneous data during eyes open and eyes closed conditions but not with very er strict erm erm constraints to the subject about what he should be doing but plotting the amplitude fluctuations er from parietal lobe er from the intra-cranial data where we know exactly how these we- er we see the strong correlation between the amplitude variability as a function of erm items entering working memory and not being er active in working memory you see quite a large qualitative similarity between the two so you might ask yourself what is the brain doing actively during rest what we are thinking if i ask you to close your eyes for a minute no- none of few of you would fall asleep most of wo- you would just continue thinking and thinking you cannot have (for your own) thinking if you don't keep your own thoughts and items where you're imagining er er you might be having a silent conversation with yourself where words that are separated by tens of seconds have to be sort of bound together . that we would like to test in different paradigms erm i'm not saying that er well one way to do so is to take erm alzheimer's patients in , whom are known to have problems maintaining er items in er in memory so what we hypothesise is that these long-range temporal correlations sort of the physiological memory of oscillations is related erm to cognitive memory so that alzheimer's patients er that have impaired working memory er will also have er impaired long-range temporal correlations in oscillations from brain regions that are involved in memory , so again we we have three minutes of eyes closed rest M-E-G data from 18 alzheimer's patients and 17 age-match controls , and what we see is er if we subtract the DFA exponents erm obtained in the alzheimer's patients from the controls then you see a large difference in the the temporal correlations in the parietal region you also see an attenuation in (up) the amplitude erm in that region but if you look at the individual er subject er and and patient data then there's a much more significant difference in the temporal structure than er than in the amplitude of the oscillations . we've then been trying to see well i have to say that these are mild alzheimer's patients because when people show up at the clinic and say i i think i have a memory problem there can be anywhere between six and ten different causes of that er of which alzheimer's disease is just one of them erm so they would like to have and the clinical diagnosis is still based on structured interviews so it would be nice if there would be some biomarkers that would help discriminate one of the different types of causes of the memory problems the only problem is that yeah in the clinic they always want clear separation erm between er healthy er subjects and and patients in order to use a biomarker as a diagnostic tool er and that is very difficult to achieve but what we do see is that if we if we plot the amplitude erm , er of of the alpha against the temporal correlation in the alpha band then we sort of get four regions of which none of the 17 erm healthy subjects have and they all they all fall in this quadrant basically it means that if no healthy subject do you have that the amplitude of the alpha band in the parietal region is lower than the mean over the rest of the head and the same goes for the DFA in no none of the healthy subjects do you see that the temporal correlations of the parietal alpha is less than it is as a mean over the rest of the head so at least you can say that we can sort of define risk zones that's highly likely to be abnormal and not just a subjective complaint if you come directly in and and say that you have a a memory problem <P:06> and well how we think that the region that we identify also it links well with with what is known about the A-D pathology we know that the- there's a (xx) of normal subjects that perform a er a memory retrieval er a memory task which shows strong activation in in the pre- (xx) a- area <COUGH> if you look at (xx) er (xx) flow or metabolism there's a deficit of metabolism in the alzheimer's patient in the in the pre- (xx) area and it's also well known that this is where alzheimer's patients have deposition of amyloid (xx) erm so somehow you can say that that by looking at the temporal structure or the abnormal temporal structure of on-going oscillations you er , you identify the same er regions of the brain with pathological or er conditions now th- this is of course well of course this is a a typographic view so it's in sensor space we cannot er say for sure that this comes from from the parietal region so we're now trying to to do source modelling on the spontaneous oscillations too and er integrate it with anatomical MRI so that we can say more firmly er (this) anatomical region actually shows these er abnormal temporal correlations </S2>
<P:05>
<S1> <NAME S2> <S2> yeah </S2> what what what method are you using for the source localisation </S1>
<S2> yeah that's er that's a good question <S1> @@ </S1> @er@ so it's not that easy well fi- first we tried to just use er spatial beamforming and what you do there is that you divide the 3-D volume into 5,000 or 4,000 er cubic centimetres or half cubic centimetres erm and then you can project the M-E-G data down to that (xx) what could possibly come there and then we do DFA on those 4,000 traits what you get i have to say that you get a pretty messy image <SS> [@@] </SS> [@@] erm so i don't think that was the way to go although i also have to say that if you look at these type of images in FMRI and PET literature they have also struggled for more than a decade to reach the status where software is so nice that you know they normalise the individual brain volume to fit to a standardised brain volume you have sort of they have agreed upon what is the smoothing algorithms we use to get these beautiful images erm so there's there's still some work to be done there it could be that the that the a better way of doing it would be to just <P:08> well , they're they're they're quite this this has been work-in-progress for 15 years people would like to to use M-E-G data for er source modelling and and it's not er it's not solved yet we didn't solve it either @@ we still hope though that we can er get into a source space , do you have a suggestion for what to use </S2>
<S1> <COUGH> i i i i think the two main well now on-line solutions would be to us- first use I-C-A and then conventional dipole modelling <S2> yeah </S2> or this er dynamic imaging of coherent sources <S2> yeah </S2> do you have o- an opinion on on DICS </S1>
<S2> no i think the problem is with DICS er that er . that i think you do need two conditions so you need for instance eyes open eyes closed which is known to var- er modulate the amplitude of of er the alpha (xx) and then you on the basis of the coherence matrix of that you can identify it er oscillations of 3-D space that are reactive to that i'm just a little bit afraid that what the what that will give you is then the alpha in the visual cortex and not so much in the pre- (xx) area so er i'm not sure that's the way to go if you want to (really s-) focus on it could be that a better approach is simply just to force a space or field to be in the parietal lobe project the data down to that and compare it to what you get if you force a spatial field which would be in the (xx) region , but otherwise i guess i'm also in favour of I-C-A er but it's quite erm that's also research in-progress how to do- deal with sometimes you find one I-C component and sometimes five erm <P:11> well what we find in the alzheimer's patient erm er studies is sort of in line with in a couple of other actually <NAME S1> has a a a paper where there you look at er long-range temporal correlations erm the intra-cranial erm or sub-dural EEG and found that in the beta band there are two strong correlations in the beta band in around the (C-G) locus as opposed to the neighbouring areas erm and there i believe you also found that the correlations were discriminating better than amplitude measures and we also have a study in depressive patients where we thought saw that the the temporal structure of beta oscillations has impaired more er than the amplitude so it's i think it's interesting if it's so that looking at the temporal structure can be often more sensitive a more sensitive measure to er cognition and and pathology than than the more classical measures erm <P:06> well that's pretty much the two studies i wanted to mention i don't know if i should move on i have , can i go on for another three minutes </S2>
<S1> sure [@@] </S1>
<S2> [@@] so this is a little bit more high-flying er when this research started it was actually er because of er a theory called s- sub-organised (xx) er from er complexity physics which claims that there's a link between spatial correlations and temporal correlations in complex systems and the brain for sure is a complex system so in the in the same way that you when looking at temporal correlations in a temporal signal can go from largely uncorrelated to strongly correlated and in-between these two extremes you have some beta stable states where you have a compromise between the uncorrelatedness and the strong correlatedness then the S-O-C theory predicts that in the spatial domain you have an equivalent of that so in the uncorrelated case if you can imagine that this would be sort of neurons and you would activate if you would activate a black neuron that it's an activatable neuron and it would propagate the activity to the neighbouring site if that neighbouring site was also a neuron that is black whereas if you hit a white region nothing no activity will propagate you can think of it as refractory or less excitable neuron now if you randomly stimulate this grid what you will see is that often nothing will happen and whenever you hit a black spot it won't propagate far because they're not very large regions contiguous regions of neurons that can be activated so by activating this grid you will actually get a very stochastically fluctuating signal with no long-lasting oscillations whereas over here where most sites are activated or can be activated almost wherever you hit erm inside a random neuron that neuron will have contact to many other neurons so you will have that the pro- that the activity propagates very far in the network and that will then correspond to the temporal signal er staying er above zero for a long time it's like a a long-lasting avalanche and in-between these two spatial regimes you have a fractal spatial pattern where you might for a period of of several activations hit a region where there's almost no neurons that can be activated but sometimes you will hit a neuron that is part of a very large contiguous region and then you will get an oscillation that lasts for very long but the difference between this one as opposed to this and this is that there's a very large distribution of possible network sizes effective network sizes whereas here you only have small ones and here you only have large ones and that we are now incorporating in in computational models of network oscillations to see what kind of spatial connectivity er network could er be the the underlying er mechanism beh- er behind the differences in the temporal fluctuations that we see . but that is also work-in-progress so er i i think i will stop there erm just a (xx) slide er many people are studying network oscillations erm people who study er humans often change tasks such as cognitive (xx) to address the cog- cognitive function erm attention and memory erm and previously have mostly used er frequency and amplitude in synchrony measures to figure out whether the oscillations were involved but now there are more traits er popping up one is long-range temporal correlations <NAME S1> has worked with cross-frequency phase correlations so he's in a sense er investigating how close frequency phase coupling can act as an additional trait to dig out erm involvement of oscillations maybe in different areas or different situations in different ways than what classical analysis would er pick up and the same goes with the in vitro line er people are studying now in many labs in vitro oscillations where you ha- can pharmacologically m- manipul- er genetically manipulate er synaptic conductances and see what their influence is on oscillations but again many of these labs are looking at the impact on the oscillations with classica- classical methods but i think there's room for much more information in this oscillatory signal than what you can er extract with classical methods so we need to develop more traits that can be sensitive to other changes or er involvements of network oscillations than can be captured with classical analysis , er yeah i i i i stop there @@ so questions if you want </S2>
<APPLAUSE>
