Incorrect use of eegfilt, epochframes must be at least 3 times the filtorder

Question

Incorrect use of eegfilt, epochframes must be at least 3 times the filtorder

Opened this issue 2 months ago · 3 comments

Description

[epochframes must be at least 3 times the filtorder. When use the estudio toolbox to create a BEST file and transform the data into phase-independent power, an error message will be displayed : epochframes must be at least 3 times the filtorder.]

Load my data which have been preprocessed by esudio

My preprocessing steps include downsampling to 500 Hz, bandpass filtering from 0.1 to 30 Hz using the 'eegfilternew' function, performing bin-based epochs, whole brain average re-referencing, ICA removes blink artifacts, and removal of bad epochs exceeding ± 100 μV.

Extract Bin-Epoched Single Trials (BEST)

Convert the data to BEST format and select 'transform into phase-independent power'

Run and Error

I clicked RUN, and then the following error appeared.

“eegfilt(): filter order is 1500. Misuse eegfilt
epochframes must be at least 3 times the filtorder.

ERROR pop_extractbest (line 291 )
filtData(:,c,:) =abs(hilbert(eegfilt(squeeze(unfiltData(:,c,:)),fs_original,bandpass_freq(1,1),bandpass_freq(1,2))')').^2; %Instantaneous power

ERROR f_EEG_extr_best_GUI/avg_run (line 654 )
[BEST,LASTCOM] = pop_extractbest( EEG , 'Bins', binselec, 'Criterion', artcritestr,'DSindex',1, ...

Error calculating UIControl callback.”

I am expecting the smooth completion of this BEST operation, but it prompts the above error. I don't know how to adjust my parameters to ensure normal operation

#### Versions


OS version	[Windos 11]
Matlab version	[2022b]
EEGLAB version	[2023.1]
ERPLAB version	[12.00]
ERPLAB Studio version	[12.00]

Answer 1 · 2024-12-05T14:36:14.000Z

Hi. Thanks for reporting this. I’m pretty sure the problem is that your epoch length is not long enough to estimate the power between 1 and 4 Hz. To estimate the power at 1 Hz, the epoch must be long enough to have multiple cycles of a 1 Hz oscillation. Typically, one would want an epoch that has at least 5 cycles of the lowest frequency in your frequency range (5 seconds in your case given that your lowest frequency is 1 Hz). This is not really a limitation of the software – it is a limitation of the underlying mathematical principles. But the software is giving you an error message because the data violate the assumptions of the math. So, I recommend re-epoching your data with a 5 second epoch length and then trying to extract the bin-epoched single trials from these epochs. Steve From: PengfeiLu1 ***@***.***> Date: Wednesday, December 4, 2024 at 10:53 PM To: ucdavis/erplab ***@***.***> Cc: Subscribed ***@***.***> Subject: [ucdavis/erplab] Incorrect use of eegfilt, epochframes must be at least 3 times the filtorder (Issue #199) Description [epochframes must be at least 3 times the filtorder. When use the estudio toolbox to create a BEST file and transform the data into phase-independent power, an error message will be displayed : epochframes must be at least 3 times the filtorder.] 1. Load my data which have been preprocessed by esudio My preprocessing steps include downsampling to 500 Hz, bandpass filtering from 0.1 to 30 Hz using the 'eegfilternew' function, performing bin-based epochs, whole brain average re-referencing, ICA removes blink artifacts, and removal of bad epochs exceeding ± 100 μV. 2. Extract Bin-Epoched Single Trials (BEST) Convert the data to BEST format and select 'transform into phase-independent power' image.png (view on web)<https://github.com/user-attachments/assets/a5063ac5-e68e-44b5-84d3-c8f1d130d2ad> 3. Run and Error I clicked RUN, and then the following error appeared. “eegfilt(): filter order is 1500. Misuse eegfilt epochframes must be at least 3 times the filtorder. ERROR pop_extractbest (line 291 ) filtData(:,c,:) =abs(hilbert(eegfilt(squeeze(unfiltData(:,c,:)),fs_original,bandpass_freq(1,1),bandpass_freq(1,2))')').^2; %Instantaneous power ERROR f_EEG_extr_best_GUI/avg_run (line 654 ) [BEST,LASTCOM] = pop_extractbest( EEG , 'Bins', binselec, 'Criterion', artcritestr,'DSindex',1, ... Error calculating UIControl callback.” I am expecting the smooth completion of this BEST operation, but it prompts the above error. I don't know how to adjust my parameters to ensure normal operation #### Versions OS version [Windos 11] Matlab version [2022b] EEGLAB version [2023.1] ERPLAB version [12.00] ERPLAB Studio version [12.00] — Reply to this email directly, view it on GitHub<#199>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/AC67QFYLKSRSFJLADXBJ2LD2D7Z4ZAVCNFSM6AAAAABTBYIWXKVHI2DSMVQWIX3LMV43ASLTON2WKOZSG4YTSNJRGQZDQNY>. You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

Answer 2 · 2024-12-05T15:08:52.000Z

Thank you very much for answering my question! I also think it's because my data length is not enough, my epoch is [-600, 900ms], but it's already the longest I can do. Due to the limitations of the experimental paradigm, I am unable to ensure that each epoch has a length of 5 seconds without overlapping. So, I still have some questions. Does this mean that my data is not suitable for conversion using this method to transform trials into phase-independent frequency-band EEG power？ I aim to achieve decoding analysis of power in the 1-4Hz frequency band. As far as I know, Estudio defaults to using FIR filters. If I replace other filters, like IIR, can I achieve a 1-4 Hz filtering conversion? Or rather, based on my current data, I cannot achieve 1Hz filtering even if I change the filter. Thank you again for your sincere help！ PengfeiLu At 2024-12-05 22:36:36, "stevenjluck" ***@***.***> wrote: Hi. Thanks for reporting this. I’m pretty sure the problem is that your epoch length is not long enough to estimate the power between 1 and 4 Hz. To estimate the power at 1 Hz, the epoch must be long enough to have multiple cycles of a 1 Hz oscillation. Typically, one would want an epoch that has at least 5 cycles of the lowest frequency in your frequency range (5 seconds in your case given that your lowest frequency is 1 Hz). This is not really a limitation of the software – it is a limitation of the underlying mathematical principles. But the software is giving you an error message because the data violate the assumptions of the math. So, I recommend re-epoching your data with a 5 second epoch length and then trying to extract the bin-epoched single trials from these epochs. Steve From: PengfeiLu1 ***@***.***> Date: Wednesday, December 4, 2024 at 10:53 PM To: ucdavis/erplab ***@***.***> Cc: Subscribed ***@***.***> Subject: [ucdavis/erplab] Incorrect use of eegfilt, epochframes must be at least 3 times the filtorder (Issue #199) Description [epochframes must be at least 3 times the filtorder. When use the estudio toolbox to create a BEST file and transform the data into phase-independent power, an error message will be displayed : epochframes must be at least 3 times the filtorder.] 1. Load my data which have been preprocessed by esudio My preprocessing steps include downsampling to 500 Hz, bandpass filtering from 0.1 to 30 Hz using the 'eegfilternew' function, performing bin-based epochs, whole brain average re-referencing, ICA removes blink artifacts, and removal of bad epochs exceeding ± 100 μV. 2. Extract Bin-Epoched Single Trials (BEST) Convert the data to BEST format and select 'transform into phase-independent power' image.png (view on web)<https://github.com/user-attachments/assets/a5063ac5-e68e-44b5-84d3-c8f1d130d2ad> 3. Run and Error I clicked RUN, and then the following error appeared. “eegfilt(): filter order is 1500. Misuse eegfilt epochframes must be at least 3 times the filtorder. ERROR pop_extractbest (line 291 ) filtData(:,c,:) =abs(hilbert(eegfilt(squeeze(unfiltData(:,c,:)),fs_original,bandpass_freq(1,1),bandpass_freq(1,2))')').^2; %Instantaneous power ERROR f_EEG_extr_best_GUI/avg_run (line 654 ) [BEST,LASTCOM] = pop_extractbest( EEG , 'Bins', binselec, 'Criterion', artcritestr,'DSindex',1, ... Error calculating UIControl callback.” I am expecting the smooth completion of this BEST operation, but it prompts the above error. I don't know how to adjust my parameters to ensure normal operation #### Versions OS version [Windos 11] Matlab version [2022b] EEGLAB version [2023.1] ERPLAB version [12.00] ERPLAB Studio version [12.00] — Reply to this email directly, view it on GitHub<#199>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/AC67QFYLKSRSFJLADXBJ2LD2D7Z4ZAVCNFSM6AAAAABTBYIWXKVHI2DSMVQWIX3LMV43ASLTON2WKOZSG4YTSNJRGQZDQNY>. You are receiving this because you are subscribed to this thread.Message ID: ***@***.***> — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: ***@***.***>

Answer 3 · 2024-12-05T18:11:51.000Z

Hi PengfeiLu. This is an issue of the fundamental nature of oscillations, not an issue of what filters are being used. In an ideal world, you need at least 2 seconds of data for the amplitude of a 1 Hz oscillation to be identified. Some filters do a better job of working close to the theoretical minimum epoch length than others, and the filter we use when we create BESTsets is very good in this regard. With a 1500 ms epoch length, you just can’t estimate power at 1 Hz. The best you could do would probably be 3 Hz. But keep in mind that filtering reduces temporal resolution, so any activity you saw at, for example, 100 ms poststimulus, would be impacted by signals that actually occurred at -500 ms and +700 ms. Steve From: PengfeiLu1 ***@***.***> Date: Thursday, December 5, 2024 at 7:09 AM To: ucdavis/erplab ***@***.***> Cc: Steven John Luck ***@***.***>, Comment ***@***.***> Subject: Re: [ucdavis/erplab] Incorrect use of eegfilt, epochframes must be at least 3 times the filtorder (Issue #199) Thank you very much for answering my question! I also think it's because my data length is not enough, my epoch is [-600, 900ms], but it's already the longest I can do. Due to the limitations of the experimental paradigm, I am unable to ensure that each epoch has a length of 5 seconds without overlapping. So, I still have some questions. Does this mean that my data is not suitable for conversion using this method to transform trials into phase-independent frequency-band EEG power？ I aim to achieve decoding analysis of power in the 1-4Hz frequency band. As far as I know, Estudio defaults to using FIR filters. If I replace other filters, like IIR, can I achieve a 1-4 Hz filtering conversion? Or rather, based on my current data, I cannot achieve 1Hz filtering even if I change the filter. Thank you again for your sincere help！ PengfeiLu At 2024-12-05 22:36:36, "stevenjluck" ***@***.***> wrote: Hi. Thanks for reporting this. I’m pretty sure the problem is that your epoch length is not long enough to estimate the power between 1 and 4 Hz. To estimate the power at 1 Hz, the epoch must be long enough to have multiple cycles of a 1 Hz oscillation. Typically, one would want an epoch that has at least 5 cycles of the lowest frequency in your frequency range (5 seconds in your case given that your lowest frequency is 1 Hz). This is not really a limitation of the software – it is a limitation of the underlying mathematical principles. But the software is giving you an error message because the data violate the assumptions of the math. So, I recommend re-epoching your data with a 5 second epoch length and then trying to extract the bin-epoched single trials from these epochs. Steve From: PengfeiLu1 ***@***.***> Date: Wednesday, December 4, 2024 at 10:53 PM To: ucdavis/erplab ***@***.***> Cc: Subscribed ***@***.***> Subject: [ucdavis/erplab] Incorrect use of eegfilt, epochframes must be at least 3 times the filtorder (Issue #199) Description [epochframes must be at least 3 times the filtorder. When use the estudio toolbox to create a BEST file and transform the data into phase-independent power, an error message will be displayed : epochframes must be at least 3 times the filtorder.] 1. Load my data which have been preprocessed by esudio My preprocessing steps include downsampling to 500 Hz, bandpass filtering from 0.1 to 30 Hz using the 'eegfilternew' function, performing bin-based epochs, whole brain average re-referencing, ICA removes blink artifacts, and removal of bad epochs exceeding ± 100 μV. 2. Extract Bin-Epoched Single Trials (BEST) Convert the data to BEST format and select 'transform into phase-independent power' image.png (view on web)<https://github.com/user-attachments/assets/a5063ac5-e68e-44b5-84d3-c8f1d130d2ad> 3. Run and Error I clicked RUN, and then the following error appeared. “eegfilt(): filter order is 1500. Misuse eegfilt epochframes must be at least 3 times the filtorder. ERROR pop_extractbest (line 291 ) filtData(:,c,:) =abs(hilbert(eegfilt(squeeze(unfiltData(:,c,:)),fs_original,bandpass_freq(1,1),bandpass_freq(1,2))')').^2; %Instantaneous power ERROR f_EEG_extr_best_GUI/avg_run (line 654 ) [BEST,LASTCOM] = pop_extractbest( EEG , 'Bins', binselec, 'Criterion', artcritestr,'DSindex',1, ... Error calculating UIControl callback.” I am expecting the smooth completion of this BEST operation, but it prompts the above error. I don't know how to adjust my parameters to ensure normal operation #### Versions OS version [Windos 11] Matlab version [2022b] EEGLAB version [2023.1] ERPLAB version [12.00] ERPLAB Studio version [12.00] — Reply to this email directly, view it on GitHub<#199>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/AC67QFYLKSRSFJLADXBJ2LD2D7Z4ZAVCNFSM6AAAAABTBYIWXKVHI2DSMVQWIX3LMV43ASLTON2WKOZSG4YTSNJRGQZDQNY>. You are receiving this because you are subscribed to this thread.Message ID: ***@***.***> — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: ***@***.***> — Reply to this email directly, view it on GitHub<#199 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/AC67QF363PGXFCM2HIYGAZT2EBUBXAVCNFSM6AAAAABTBYIWXKVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDKMRQGU3TSNBSGM>. You are receiving this because you commented.Message ID: ***@***.***>