[color=#000000]Hi Sascha & Pravesh[/color]
[color=#000000]Just to confirm, yes, the average BOLD signal at each voxel separately for each session is always removed during the denoising step (first as a always-on constant term to the regression equation, and second as a result of band-pass filtering).[/color]
[color=#000000]Hope this helps[/color]
[color=#000000]Alfonso[/color]
[i]Originally posted by Pravesh Parekh:[/i][quote][color=#000000]Hi Sascha,[/color]
[color=#000000]Yes, I can confirm that for my dataset too the value is close to zero (order of -9). I think it might be because the unit of analysis is percent signal change. Consider the following:[/color]
[color=#000000]At a ROI level, PSC is calculated by scaling the time series with the average time series in that ROI. Using the same logic, for voxel level, the scaling would be done using the mean time series of that voxel. Therefore, for a given x,y,z voxel:[/color]
[color=#000000]PSC = ((voxel_time_series - mean(voxel_time_series))/mean(voxel_time_series))*100;[/color]
[color=#000000]The average time series that we are calculating from the denoised time series is essentially mean PSC rather than mean time series as such. Since its scaled to its own mean, we are seeing a value close to zero. Does that make sense?[/color]
[color=#000000]Best[/color]
[color=#000000]Pravesh[/color]
[i]Originally posted by Sascha Froelich:[/i][quote]Dear Pravesh,
That's right, I am talking about the niftiDATA_Subjectxxx_Condition000.nii file in the preprocessing folder.
I calculated the time-average for every voxel and then looked for the largest absolute value of all the time averages, which was at the order fo 10^(-8) or sometthing similarly small:[quote]%data is a 91x109x91x300 matrix containing the imaging data (300 timepoints)
[rows,columns,slices,timepoints]=size(data);
meandata=zeros(91,109,91);
for t=1:timepoints
meandata(:,:,:)=meandata(:,:,:)+data(:,:,:,t)/timepoints;
end[/quote][quote]>>max(max(max(abs(meandata))))
ans =
2.0117e-08[/quote][/quote][/quote]
[color=#000000]Just to confirm, yes, the average BOLD signal at each voxel separately for each session is always removed during the denoising step (first as a always-on constant term to the regression equation, and second as a result of band-pass filtering).[/color]
[color=#000000]Hope this helps[/color]
[color=#000000]Alfonso[/color]
[i]Originally posted by Pravesh Parekh:[/i][quote][color=#000000]Hi Sascha,[/color]
[color=#000000]Yes, I can confirm that for my dataset too the value is close to zero (order of -9). I think it might be because the unit of analysis is percent signal change. Consider the following:[/color]
[color=#000000]At a ROI level, PSC is calculated by scaling the time series with the average time series in that ROI. Using the same logic, for voxel level, the scaling would be done using the mean time series of that voxel. Therefore, for a given x,y,z voxel:[/color]
[color=#000000]PSC = ((voxel_time_series - mean(voxel_time_series))/mean(voxel_time_series))*100;[/color]
[color=#000000]The average time series that we are calculating from the denoised time series is essentially mean PSC rather than mean time series as such. Since its scaled to its own mean, we are seeing a value close to zero. Does that make sense?[/color]
[color=#000000]Best[/color]
[color=#000000]Pravesh[/color]
[i]Originally posted by Sascha Froelich:[/i][quote]Dear Pravesh,
That's right, I am talking about the niftiDATA_Subjectxxx_Condition000.nii file in the preprocessing folder.
I calculated the time-average for every voxel and then looked for the largest absolute value of all the time averages, which was at the order fo 10^(-8) or sometthing similarly small:[quote]%data is a 91x109x91x300 matrix containing the imaging data (300 timepoints)
[rows,columns,slices,timepoints]=size(data);
meandata=zeros(91,109,91);
for t=1:timepoints
meandata(:,:,:)=meandata(:,:,:)+data(:,:,:,t)/timepoints;
end[/quote][quote]>>max(max(max(abs(meandata))))
ans =
2.0117e-08[/quote][/quote][/quote]