I will just give you how to do the EEGLAB and ERPLAB analysis with an MMN data. So, I will be giving you a brief introduction about what is this MMN and how it looks and everything. (Refer Slide Time: 0:48) So, MMN it means it is called the mismatch negativity. It is actually a negative component that is being obtained were during the analysis, during the ERP analysis, EEG analysis. So, we in this we are giving two types of stimulus, two stimulus are given at the stretch like standard 500 hertz and a deviant of 1000 hertz and that too it is in it is repeated, it is repeated in a particular fashion. For example, it will be always in this fashion beep beep beep boop, like that it will
So, it almost every time it will be in this fashion because it should be in a repetitive form, that is when only then that is when we can find out when the deviant response is obtained. So, the brain what it does is it keeps the standard, the standard it will think that this is the how the sound will come and suddenly when there is a difference in this, in the tone that is when the deviant response is obtained, there is a different negative component that comes into part, that is called as this MMN, that mismatch negativity that is why it is known as mismatch negativity. This is also similar to that of the auditory stimulus, auditory evoke potential itself because even in this also there is an auditory because it is an auditory stimulus, we are giving there is an auditory cortex is, it is involved in this particular MMN ERPs also. So, usually, this MMN if you see, if it will be always between a 150 to 300 millisecond, in this particular range the MMN should be seen every time. So, it starts with the deviant stimulus, it starts soon after there is a deviant stimulus obtained, there will be this MMN response being obtained. Next, the so experimental details, it is again recorded from the neuro scan system, a wet electrode system.
So, here we have it is a 64 channel again and then we have the stimulus as I told you, it should be this S it will be standards and for a deviant and followed by 4 standards and 1 deviant like that this will be repeated for about 100 loops or something like that. So, it we will get a huge and the time range of this MMN, it would maybe takes about 10 to 15 minutes like that. So, then only we will be able to average all the ERPs and everything and to obtain the particular MMN. And even in this also we have all the filtering range, we have we will do all the filters, we will be incorporating the notch filter, the bandpass filter, and everything and the artifact rejection is again from minus 1000, 100 to plus 100 microvolts. So, even in this experiment what happens, we have a longer time range, we having more, mean it is taking 15 minutes of where we cannot steadily or stationary like that we cannot sit for long, so obviously there will a lot of artifacts been produced in this also. So, that is why we have to have the artefact rejection to be done in all the ERPs. Then, and this mainly the MMN it is being obtained in the frontal, right frontal part of the brain it will be obtained. So, for example, I will just show you in the next pictures. So, here this blue part over here that is the place where the MMN response is maximum because here is the place where you have the primary and the non-primary auditory cortex and all are present in this part of the brain.
So, here in this place only, the MMN is being obtained. So, here in the first figure as I showed, here we have the there are 3 standards followed by the deviant stimulus, so what happens is, the all the standard we can together average it and obtain one form and all the deviants together we can average it and produce this waveform. The thin black line over here is the standard stimuluses waveform and the thick line over here that is the deviant stimulus. So, that is the waveform obtained for the deviant stimulus. Now, what we have to do is, always when this MMN, it should be the difference between, there will be the MMN is not, it is not just the deviant stimulus or the standard stimulus, it is none of that, it is the difference between the deviant and the standard. So, when we so, when we have his ERP analysis being done, we have to separately being, we have to create two different bins for the standard as well as the deviant and then we have to subtract the, we have to do a bin operation so that there will be a subtraction taking place between the deviant and the standard, so then only we will get the proper MMN result.
So, here in this case, here it show in this figure it is shown and the left figure the it is shown as a negative down or a positive up as shown. So, here if you see it is a perfect peak at the 200 millisecond and here we can see that it is between 150 to 200, you will have 100-200, there is a peak. This is a two different pictures I have taken just to show which part of the brain is being activated in this case and what is the different stimuli, how does the stimuli looks, for that we are using this. So, always the standard will be a 500 hertz tone and 1000 hertz tone. Now, there are a various types of MMN also like there can be duration MMN. For example, now from one deviant from the standard to deviant if you have, we can change the timeframe. Now, usually sometime, mostly people take a 500-millisecond difference or a 2 hertz of frequency they will take, in that range they will experiment. So, sometimes they will take a different timeframe also. So, there are different types of MMN also been there like this duration MMN and the delta MMN, like that there are various types of MMN as also there. We can modify the experimental protocols according to what we want to find out. So, different types of this one will have obviously will have different kinds of result also. So, this is about the MMN pictorial representation of how it looks.
And so how does and I will just go brief about how to do this MMN demonstration in EEGLAB and ERPLAB. So, again as I told you, first we have to have the format according to the EEG thing, I mean how whichever EEG format we have, according to that we have get and then he will import the EEG data and then we do the notch filtering and followed by we just check what which trigger refers to the standard and which triggers refer to the deviant we have to check that before doing any analysis. And then do the create the event list for the standard and deviant separately and then we have to do the artifact rejection, all the band passing filters and everything and then finally we have to epoch the MMN and we have to, another important step apart from the EEP is that we have to do the bin operation in this in this thing because we wanted the MMN is the difference between the deviant and the standard. So, we have to do that particular step in the ERPLAB analysis before doing any other. Otherwise, we will just directly if you do not do the bin operation, we can just visualize the standard alone and the deviant alone but we would not be able to visualize how the MMN looks. Even the channel operations also can be done for this particular because as I told you only the right side frontal-central in that particular area only the brain is being activated. So, only those electrodes and all can be together averaged and channel operation can be done, averaged together in that part and then we can obtain a final MMN waveform. So, this is about the, this is about the total, procedural of procedures of what we have to do in the EEGLAB.
And the MMN if you have to say about, this is proper like this is Naatanen is the main person who done lot of research on this MMN, various other different types of MMN, auditory, visual or even this duration MMN as I told you, different types of MMN and analysis has been performed by this Naatanen and group. So, these are the best references you can give for MMN. You can have this as the base paper for doing any MMN analysis. So, next I will just go about demonstrating how does the MMN EEGLAB, ERPLAB analysis we can do using the Matlab source, so thank you.
Hello everyone. So, we have doing the MMN we just give a introduction what, how this MMN looks and everything. So, I will just give you how to demo, how to proceed with the EEGLAB and ERPLAB analysis with MMN data. It is a similar steps only but some difference in the event list and all will be there. So, I will just demonstrate as such. So, here we have, we will just have the, I will use the graphical user interface method itself I will just show you how to create the scripts at the end. So, here we have the EEGLAB being running, accordingly we have various formats.
Now, in this experiment, I will use another format of EEG instead of the CNT format, I will use Enobio format of EEG, a min of the MMN file. Like I will take a NEDF file for example. So, here I will just import the MMN data. So, this is this neuro electrics is actually, these are the plugins that are there in all the EEGLAB website itself. You can directly go there and download it and just you have to paste it in the plugins as I have mentioned in the previous demonstration, we have to just paste it over the plugins of that EEGLAB. So, here we have these are the 2 formats that is being obtained for open BC, the Enobio and all. So, NEDA for EEG. So, I will take this NEDA format over here and I will just take up MMN data, this is the MMN data I am taking over here and here we can have, we can load the channel. So, as I told you in the previous demonstration, I mentioned that we can there be channel locations which where we have to manually do it or sometimes in the in file itself there will be these channel locations present. So, this is one such format in wherein the channel locations are also there along with the continuous data. So, this is a one so just we will load the channel locations and then give an ok.
So, here it reads all the functions, and then it will give you the, it will read the locations and everything and I will just name it as MMN, MMN demo, ok. So, here it open as this. So, here in this data format, they are having only 32 channels and then the epochs are from, there is only one epoch because we did not do any epoching yet. So, it is safe for here it is like 1, 568 seconds it is lasting which means that it is more than 15 minutes like that. So, this is the time length of the experiment over here. Here already the channel locations are being included and the events, I will just show you the events and here the sampling rate is 500. We can change this sampling rate I will discuss changing the sampling rate in the wavelet analysis section. So, here I will just show you how the data looks. So, I will just change the so that we get all the channels. So these are there are only 30, 32 channels only are there, so here if you see there are these are the standards and there is a 1 deviant and there is a standard and then a deviant, standard and a deviant. So, this is how it a and it is a repetitive format. It is not like here there are 6 and here there are 5, not like that. Here all are 6 and 1 deviant, like that it is. This is how it should, the MMN data should look like.
So, before doing any analysis further, we just do a pre-processing step of the filtering. So, here we take again a notch filter as usual. And put a 50. So, this is the usual step as I told you. So, if after look, after doing the filtering, maybe the data will be looking even better. So, I will just show you the thing. So, here see the proper, the after filtering itself, there is a proper seen of the all the channels have been shown properly and we can change the number window length or like we can make it 20, 25 like that. And also we can visualize the channels. For example, now I want only 2-4-5 channels to be seen at a stretch means we can do that also. So, we can change the amplitude accordingly like I wanted to be a very less amplitude. So, we will get a proper visualization of how it looks like this. So, this is just a pre-processing step of notch filter I said ahead. Before that, after this we have to do create the event list. So, here we have two events as I told you, one is there to standard event and the deviant, those are the two events that are been there. So, first, we have to do the event list, so we have to create the event list accordingly. And then we have to do how it looks.
So, first, there are there is one event wherein which is actually a 500, which is a 500 hertz and we just name it as the first event and it is called as a standard. So, let it be where we will name it as a standard. And then we update. So, this is one, this is one event. For the next event, we will make it 2 because there are 1 and 2 events, so here this is the 1000 hertz and then we have we may name it as the second bin and here we will name the bin description as deviant we have the deviant stimulus trigger. So, here we do, so this is how the event list has been created. And then we can save this as MMN, however, you want you can save it in that format. So, I just save it as MMN total. So, this is how the event list has been created. Now what happens is all the total, all the events in that, all these 1s and 2s will be replaced by bin 1, bin2 with these standards and deviants and everything. So, we will just apply that.
So, while doing it will create all the bins. So, here is the pre-processing how it looks in the command window, so here it creates the two events for 1500 hertz and for 1000 hertz for the standard and deviant specifically and then we name it in the form of code labels and we just apply it and it will create an event list, so we just do that. So in this way we have made the event list. Now, as I told you we have to do the epoching as well. So, as I told everywhere we have to do that epoching step in all the ERPs. So, what happens we have to do the epoching, we have to do the bin based epoch. So, here also we wanted we just may keep up baseline correction of minus 50 because we just do not want as soon as how the as soon as the trigger starts, how the waveform changes we do not want to know, we just want to have a small baseline correction of how it looks before the trigger starts. So, we keep minus 50. Usually, how much ever range we take, 10 percent of that we have to keep it in as a baseline correction. So, if you are taking for 1000 seconds means, 10 percent of that 100 milliseconds of baseline correction you have to keep, so I just keep a default as minus 50 for any AEP any ERPs as such. So, I take as I told you for MMN, it is from minus it is like 150 to 300 in that range the MMN will be there. So, that is why we will keep a range of about 400 or 450 like that so that we will get a proper, we can visualize all the MMNs properly, for that purpose only we do this. So, I will just run it.
So, what happens is it will run in such a manner that it will get epoched in that particular frame time range. So, the bins have been created. So, if you see here there are 500, 700 events which means total of all the standards and deviants together there are 700 events. In that 700 events, we have made the 700 epochs also which starts from minus 50 to 500, to 450 millisecond. So, this is how we do the epoching. After this, I will just show you how this epoch looks. So, here we have seen all this, these are the bin 1s all the standards and followed by the a deviant, standard and a deviant like this. So, I will just show a for about 12. So, over here we have, so here if you see there are deviant followed by the standards, a deviant and the standards like that. So, this is how the MMN should look like. Now, what happens is when you do the averaging, only these bin 2s will together it will get average and all the bin 1s will together get averaged. So, like that we will get two averages of both the bins. Next I will just show you how to do the artifact rejection. So, artifact rejection as I told it should be always the voltage threshold. Now, in this we cannot take all the channels because here in this there is no separate trigger channel is not there. In the previous demonstration as I showed in the CNT, the trigger channels was separate and other EEG, EMG channels and all will separate so we just excluded that we did only the other electrodes together. But in this all the channels, all the EMGs, all the triggers everything are in the channels itself. So, for that reason what we have to do is we have to just select few the actual areas where for example only the frontal, frontal electrodes or the frontal central electrodes and all of this the place where MMN is maximum seen.
So, we just take those electrodes alone like all these, these are the frontal electrodes and these are some parietal electrodes that are there. So, only these few electrodes, only those electrodes we will be taking I will just take all these electrodes 4, 5, 9 electrodes as such and only these only for these channels I will do the artefact rejection, not the others because others are not mean the MMN has more specifically seen in these electrodes rather than the others. So, I will just take that and I accept it. So, if you go in the back and see, you will see how many of them are being rejected and how many of them are being accepted. So, over here if you see like it is like a 50-50. So, if I had selected the whole of EEG all the channels, if all the channels would have been rejected, all the epochs would have been rejected. So, here there are 2 bins, in that some 100 of them are for standards and the remaining were for the deviants. So, then we will just do this, so here we get the accept these many of them are being accepted. So, we have to get at updated in our this one also. So, we just do that. Then here in this yellows, the yellow one which is being highlighted in yellow, they are the ones which are is having the those are going to be rejected. So, we just update these marks. And then we go as I showed in the previous demo also like we have to go to these tools and inspect and then we just reject the marked trials.
So, all these yellow marked try, these are the all yellow all these highlighted ones are the marked one and that will get rejected. So, we will get a see over here we can see that out of 700, these many trials have would been rejected, has been rejected. So, we will get a new dataset. So that is how it looks. (Refer Slide Time: 14:34) And after this our regular step is filtering. So, we will do a bandpass filter. After doing the bandpass filter, we will do the remaining other averaging ERPs and everything. So, it should be always 1 to 30, we can change it. Here we have the order, we can change it or we can even change the high passband and the low passband and everything. So, we just apply it and then we will get the averaged, the filtered dataset. So, we have one now, as I told you there are 2 bins now we have a bin 1 and a bin 2. So, bin 1 is for the standard and bin 2 is for the deviant. So, we will further do the averaging of those two bins. So, here in this seventh dataset after doing all the steps filtering, event list, epoching, artefact rejection and then filter, rejecting the rejected artefacts and then finally filter, we will get a just these many, this 344, only these many channels, I mean only these many epochs have been created, of that few of them are for bin 1s and few are from bin 2 and they will be separately averaged. So, we will be doing that step in this ERP set average.
So, here it is just a pop up saying that about the emerging merging. So, we will just name the MMN. So, here we have the MMN demo I will just save it in this format. I will just show you how it looks with just two events, two bins it shows actually. So, here we have the MMN demo is been generated. So these are the till now whatever we were doing it was being data, this dataset was being obtained over here. Next I will just show a plot of this. In this plot we do not want much of these channels, I will just take a few of them like the central and frontal part, so actually I will show everything. And then I will just show a topographic manner how it looks. So, in the topographic sense if you say and here also we should take since it is a MMN like it is a mismatch negativity, so it is always should be a negative up but we have not made the MMN yet. This is just the two bins like the standard and the deviant separately we have epoched and kept, averaged and kept.
So, here the red ones would be the, so if you see over here the red ones will are the deviants and the black, the black line over here, the black waveform is the standard. So, this is how it looks. Now, we have to do the subtraction between deviant and standard then only you will get the proper MMN data. So, I will just do that operation now. So, this is just to show that this is how it looks with just standard and deviant together it looks in this manner. I will show you how to do the bin operation. So, here we have the ERP operations, in this, we have to take a bin, the bin operation function. So, here we can do this is the two bin we have, we have to do the bin 2 minus bin 1. So, it will we can make it either in a in the current dataset itself we can d or we can create a new dataset also. So, I will just do new data set. So, this is how it does, b2 minus b1 and I just run it. So, it will give you the bin. So, I just write it as a bin operate, bin operation and I just save it and I will show you how does it look after the do, after doing the bin operation how it looks. So, this is how the bin operation is being done and then I will just show you the bins.
So, over here if you see this part, this is the actual part of the brain wherein the proper MMN should be looking, we should get. So, I will just show you this particular electrode. So, over here if you see it is this particular hump over here, this particular negative over here, this is called as the proper MMN. So, it should be actually from minus 50 to 300, in this range only but we are obtaining this in this range over here, this particular peak, this part this is called as this is called as the MMN for us. So, we can check what is the data points also, how it looks and everything. So, this is the classic MMN response which is being obtained between so minus like 152-300 in this range it is being obtained this point, this is the place where it is. So, we can what we can do is, we can directly do all the channel operations also. Like we can just take the average of only these particular this area as I showed in the p presentation, it was that blue area was the place where the MMN should be there which is in this part, this is the actual area where it should be there. So, we can do the channel operation in that way also.
So, I have already that being done. So, here if you do the channel operation, so here I have done the channel operation over here like we have to do for the front, we h to select which channels is for this front frontal central left side, the right side of the brain, so we will just select all them, all of them and just average it. This is an equation that I made earlier, so we can just select and add and do the averaging. And if you run you will get a channel operated one. So, just do a, we are chopping it to a particular part of the brain. So, I will just do that and I will show you. There will be just one channel and one bin which just shows one of the ERP. One minute, I will just make it a bigger, we will make it negative up always because as I told you it is a negative peak. So, here this is this particular this is called as the MMN here. From here to here, that is the proper MMN so it starts from about, it starts from about 170 and it ends in this particular point. So, if you do all these operations we will be able to clearly see the MMN response. So, this part and this part is not necessary for an MMN study. So, only we have to concentrate from this part to this part that is the place where you have to find the MMN response. So, after, doing the channel operation and the bin operation, we will be able to get the proper MMN result. Now for the same similar way there will be a lot of subjects, for many subjects we will do this similar data analysis. So, if you do all those analysis together and we can perform a grand average, we will be getting the statistical analysis we can do the T test and other Enova and everything and all and we can see how the standard, so how the SEM look like and all we can find out.
Another important step I had to mention about is the scripting as I told you. Here, we can whatever steps we have being doing till now, it will be there as a history, it will be there in the EEG history. So, if we just do eegh if you put if we just put this eegh command over here, we will get what all other codes which we have been obtained for all the codes required to do all the other analysis. For example, for the now, this is the loading the dataset and then we can just do we did the notch filtering and then here this is the event list creation and then we did the epoching. So, like that all the, how a sequentially how we had done, we will get the proper scripts. So, we will we can just copy this and paste it accordingly in our editor or every anywhere and then we can just name it accordingly and you need to just for multiple data analysis and all, we can just use this particular dataset. The only thing we have to just change the subject name or the name of the data file, accordingly we have to change and obtain the result. So, this is just a step that I missed in the previous demonstration. So, this is how we do the MMN analysis.
So, as then as I told you this is the step where we can do the grand averages of you have multiple subjects, all the datas you have together, we can collect all these ERPs together and finally do a grand averages together. But the main important step about these grand averages is it, is that we have to keep the data, the processing constant like it should be all the filters should be the same, all the epoching, the time range everything should be the same. If you are not do, a different time ranged one together for epoching, that is not possible. Even the sampling rate also should be the same for both the for doing the grand averages. So, there that is all about the MMN. So, this is the MMN data, it is classic MMN data. So, the next demonstration or the next presentation I would be giving is about the P300 and how to do its demonstration and analysis.
Log in to save your progress and obtain a certificate in Alison’s free An Introduction to Brain-Computer Interfaces online course
Sign up to save your progress and obtain a certificate in Alison’s free An Introduction to Brain-Computer Interfaces online course
Please enter you email address and we will mail you a link to reset your password.