Dear List
I have also noticed this issue in 30/30 datasets.
Looking at the code I'm not sure why it doesn't occur more often. I note that the reference volume gets subtracted from the time series so that entry become 0. When the differences are then calculated to "remove slow trends" having a 0 in the time series causes an artificial "spike". I think it probably happens more in datasets where there may be a bit more movement on average, causing the reference image and the image next to it to have an exaggerated difference.
Is there any reason the reference image has to be the one in the middle of the time series? Could it be the first image (after the dummy scans)? This ends up causing a loss to only the first image.
Darren
|