listmode files can start with non-zero time-tag

[KrisThielemans]

Timing is currently in terms of the value of the time tag. For Siemens files, the first time tag is usually zero, but in ~1% of the cases not. GE files never start with zero. It would therefore be better to have timing relative to the value of the first time tag. Current work-around: use list_lm_events to find first time tag and create initial “skip-frame” in the frame-defs file.

[KrisThielemans]

example script:

first_tag=`list_lm_events --time 1 --coincidence 0 --num-events-to-list 1 $list_filename`
echo "0 $first_tag" > $fdef_filename
# now have the rest of the fdef info, e.g.
1 $frame_duration

[AnderBiguri]

@KrisThielemans How is the skip frame defined then? will STIR know to ignore a frame that contains zero events in the listmode, or will it actually attempt to recon/process a PET image with all zeroes in the sinos?

[KrisThielemans]

an fdef with a line 0 100 will make LmToProjData et al skip 100 seconds. In this case, there are actually no data in those first $first_tag seconds, so it won't do anything, except make sure that all the rest is ok.

[AnderBiguri]

Ah, my question was more: Does it skip it because it skips whatever is in the first line? Skips whatever starts with 0? or checks if there are events, and skips if there are no events? What is the logic?

[KrisThielemans]

it advances the start of the next time frame without storing data because of the first 0

STIR/src/include/stir/TimeFrameDefinitions.h

Lines 65 to 67 in aae4c92

	This class in fact allows an extension of the above. Setting
	\a num_frames_of_this_duration to 0 allows skipping
	a time period of the corresponding \a duration_in_secs.

the fdef doesn't check events itself at all.
LmToProjData and other loops only store data that are within the time-frames asked for (but it assumes that the clock starts at zero, which it doesn't for GE data, so we have to "skip" the first time until the start of the GE clock)

[AnderBiguri]

In any case, a more complete example script would be:

first_tag=`list_lm_events --time 1 --coincidence 0 --num-events-to-list 1 $list_filename | tr -d '\n' | awk '{print $NF}'`
echo "0 $first_tag" > $fdef_filename

Just removes newlines and grabs the last word in the output of list_lm_events, as its a bit verbose.

[AnderBiguri]

@KrisThielemans your proposed solution does not seem to work for GE data. the first prompt in my example dataset is at Time 7467, but when I do list_lm_events --time 1 --coincidence 0 --num-events-to-list 1 $list_filename, each single listed with +1 in that number. If I use 7467 to define the start time (by ignoring that amount of seconds) then it generates empty sinograms.... Only if I make a fdef with content 1 10000000 will the code generate non-empty sinograms for GE data.

Not sure if this is a problem with how list_lm_events reads GE data, with different scale of the numbers (ms vs s?) or something else, but just in case, letting you know that the example does not work for GE. I can't really test this because my n_samples=1