Drop channels after bipolar reference

Hi, I have a technical problem. I was analyzing SEEG. Normally, we should drop channels out of the brain. After doing bipolar reference, the signal doesn’t simply present the original channel. Like the signal of A2 becomes A2-A1. So how do we process this situation when dropping channels out of the brain.

Thank you

Hello, I don’t fully understand your question. So you’re adding a bipolar-referenced channel. Then what? You’re free to remove the original channel. I’m not sure where the issue is here? Could you please elaborate?


Hi, sorry, my bad. SEEG is one kind of iEEG, so when analyzing SEEG, we need to drop the channels out of the brain, keep the channels which are inside the brain. While doing bipolar, each channel was re-referenced to its adjacent channel on the same electrode shaft. Like in shaft A, we have A1, A2 A3, A4, A5 and A6, after doing bipolar, we’ll have A2-A1, A3-A2, A4-A3, A5-A4 and A6-A5. Let’s say A2 and A3 are out of the brain, normally we should drop A2 and A3. But after bipolar, A2 becomes A3-A2, A3 becomes A4-A3, so I’m not sure if it’s correct to drop (A3-A2) and (A4-A3).

mne.set_bipolar_reference() has a parameter drop_refs, which defaults to True and removes anode and cathode. So it’s not that A2 becomes A3-A2; instead, A2 is removed from your data, unless you set drop_refs=False.

So, basically, I don’t need to drop the re-referenced channels because of the position, unless the quality of the signal is pretty bad.

I would say yes, that’s right.

Still makes me wonder why we set the locations in this manner…

1 Like

Sorry, what do you mean " set the locations in this manner"?

It seems like this question is asking “what is the right thing to do” rather than “how do I do this in MNE-Python”. My instict would be that if there are electrodes outside the brain, they should be dropped before re-referencing to adjacent electrodes on the same shaft. But that’s just based on general principles, I don’t know what is typical for sEEG studies. Pinging @adam2392 who might know more.

I’m not sure which location the newly-created virtual electrode should have. It seems we assign to it either the anode or cathode location (I didn’t check which of the two we pick). But I don’t know if that’s actually a good idea. Maybe it is :thinking: I guess it’s more of a philosophical question :slight_smile:

Normally for SEEG if I do bipolar I first remove all channels I don’t want to consider and then I do bipolar referencing.

1 Like

So you mean like, if A2 and A4 are out of the brain(delete A2 and A4 in A1 A2 A3 A4 A5 A6), you first delete A2 and A4 channels, then you do bipolar and get A3-A1, A5-A3, A6-A5 according to (A1 A3 A5 A6)


The whole purpose of bipolar referencing is to remove common noise in spatially close regions. If you bipolar it with a non neural signal, just introducing additional noise.

Thank you all for your help!!!Really appreciate that.

So that means I should remove all the bad or noisy channels before I do bipolar? Or could I do bipolar first and then remove the bad bipolar channels? Or both ways are acceptable?

Hello @SuperX-666 and welcome to the forum!

Often it is difficult to tell whether data produced by one sensor is good or warrants an exclusion of this channel before it has been properly referenced. So I suppose I would always reference first and only then decide whether or not to remove the channel.

You could reference the answer in the brainstorm community. Basically, if it’s not the quality problem of SEEG electrodes, the bad signal is very rare in SEEG for its high SNR. If the contact is not in the brain, according to the propagation effect, you still could use the signal to do bipolar, and frankly, the MNI coordinate of contact is the coordinate of the centroid, in the real world, the contact has a diameter. So if it’s not the tail of the shaft, basically it’s OK to use the contact that may look like not in the brain to do bipolar.

Thank you for your answer.
In my SEEG data, I have seen contacts that are contaminated by 50Hz but hardly seen after I do bipolar. That’s why I have this question. I wonder to do notch filter before or after bipolar?