You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I open this issue so we can discuss the point raised by PR #115; especially when remove superficial noise using short-distance-channel.
As I understand, the main pipeline currently is something like :
Import data
Motion correction
Detect bad channels
MBLL or doD removing SSR
IIR filtering
Block average or GLM
MEM or MNE (if computed dOD in 4) [done before 6 if GLM]
So if I understand it well, we should separate the superficial noise regression from the dOD computation and the pipeline would become :
In the case of block averaging :
Import data
2.a Detect bad channels
2.b Motion correction
Compute dOD
4.a Regress out signal from short-distance channels (or external measurement such as accelerometers )
4.b IIR filtering
MBLL or step 7.
Block average or GLM
MEM or MNE (if skipped step 5)
With the exception that in case of GLM, you have to skype step 4.a and regress those signals out during GLM.
Questions:
a. should we do motion correction after dOD computation? The advantage is that we don't have to worry about negative value induced by the process and remove the offset introduced in 2d34b50 that was making us worry during the mini-course presentation.
b. Line 499 in the computation of MBLL it is written, about the extinction coefficient, "%TODO: use same values as Alexis". Do you have any information concerning this TODO?
Changes :
Therefore here is the change we have to make :
1/ we can remove MBLL_SRR and dOD_SRR
2/ Add a process to regress out signal from the short-distance channel (can be change latter to also regress out signal from the accelerometer)
3/ Change MBLL so it akes dOD as input.
Am I correct?
Best regards,
Edouard
The text was updated successfully, but these errors were encountered:
Hi,
I open this issue so we can discuss the point raised by PR #115; especially when remove superficial noise using short-distance-channel.
As I understand, the main pipeline currently is something like :
So if I understand it well, we should separate the superficial noise regression from the dOD computation and the pipeline would become :
In the case of block averaging :
2.a Detect bad channels
2.b Motion correction
4.a Regress out signal from short-distance channels (or external measurement such as accelerometers )
4.b IIR filtering
With the exception that in case of GLM, you have to skype step 4.a and regress those signals out during GLM.
a. should we do motion correction after dOD computation? The advantage is that we don't have to worry about negative value induced by the process and remove the offset introduced in 2d34b50 that was making us worry during the mini-course presentation.
b. Line 499 in the computation of MBLL it is written, about the extinction coefficient, "%TODO: use same values as Alexis". Do you have any information concerning this TODO?
Therefore here is the change we have to make :
1/ we can remove MBLL_SRR and dOD_SRR
2/ Add a process to regress out signal from the short-distance channel (can be change latter to also regress out signal from the accelerometer)
3/ Change MBLL so it akes dOD as input.
Am I correct?
Best regards,
Edouard
The text was updated successfully, but these errors were encountered: