Time differences between MR task trigger and DICOM Acq time

Overview

This repository demonstrates pulling task onset times from two independent sources, the task display computer and the scan computer, to check for mismatches as a result of erroneously missing TTL TR triggers (= sent to start the task).

We can combine the two sources useing participant ID, session date, task name, and run number to explore timing differences.

Unfortunately, we cannot directly compare onset times because the computer clocks are not synced. But that can be tackled different ways

1. With enough display onset + MR onset pairs, we can model the offset and dift between the clocks.
2. For sessions with multiple back-to-back tasks or runs, for example the AntiSaccade task herein, we can compare the MR-display offset from run1 to run2 to check the task start time onsets. If TTL was not sent when expected, this difference of differences will be non-zero and greater than the TR of the task acquisition.
3. deconvolve high HRF responses in fMRI task data with a TENT model to measure peak response against expected time

See Makefile for recipes for both.

Notes on timing

Computer clocks drift due to frequency error

Even if the systematic error of a clock model is known, the clock will never be perfect. This is because the frequency varies over time, mostly influenced by temperature, ... or magnetic fields. ... oscillator’s correction value increases by about 11 PPM.. .. 12 PPM correspond to one second per day

On the scan side, first dicom's Acquisition Time is the closest to the pulse send time (see mr_time.bash). neurostars, snug-discuss

(0008,0032) Acquisition Time → MARS timestamp MHD.DICOM.AcquisitionTime is defined by the timestamp of the first ‘imaging’ Readout of a certain slice (Mdh.isImagingScan(), Mdh.getTimeStamp()).

On the task side,

• E-Prime logs session start (task loaded, SessionSTartDateTimeUtc) with only second precision (?) and hopefully first = translated scanner TR pulse like WaitForScanner.OnsetTime in milliseconds. See extract_eprime.pl.
• PsychoPy can record flip time in unix epoch. See lncdtask_display_time.bash.
• In psychtoolbox, on Windows, GetSecs() is "number of seconds since system start up" and will need an additional timestamp to covert events to time. Maybe file modification time?

Modeling drift

• MR clock start time for all epfid2d1 (bold EPI) are in txt/mr_times_p2.tsv.
• an example for parsing eprime logs is in extract_eprime.pl.
• pairing the two (WIP) in merge.R. Currently (20241223), model_time.R has the most insight into the clock drift.

MR time is moved into UTC timezone to avoid daylight savings time jumps. Still, the drift has discrete jumps likely from one clock being adjusted manually.

Example session diff

script	desc
lncdtask_display_time.bash	extract "EPrime" PC task start time from log
mr_time.bash	extract acquisition start time from first dicom in series
merge_luna_anti.R	merge times and explore the onset differences

output file	desc
txt/combined_tdiff.csv	diff of differences, see merge.R
txt/combined_anti_times.csv	row per visit, merged times ready for inspecting
txt/luna/anti_task_display.tsv	= trigger recieved times from task/display PC
txt/luna/anti_task_mr.tsv	DICOM Acq Time headers

Pull repo data (Git Large File Storage)

using git-lfs to store csv and png files.

git clone https://github.com/NPACore/p2-timing-debug
cd p2-timing-debug
git lfs install
git lfs checkout

Stats

tdiff <- read.csv('txt/combined_tdiff.csv')
summary(tdiff$tdiff)
     Min.   1st Qu.    Median      Mean   3rd Qu.      Max.
-338.9225   -0.0135    0.2300   -4.0310    0.5670   72.8050

More than 1 seconds off

times <- read.csv('txt/combined_mr_task_times.csv')
tdiff <- read.csv('txt/combined_tdiff.csv')
tdiff %>%
 filter(abs(tdiff)>1) %>%
 transmute(sesid=paste(luanid,gsub('-','',vdate),sep="_"),diffdiff=tdiff) %>%
 merge(times, by="sesid") %>%
 head

           sesid diffdiff run          acqtime_mr        acqtime_task    tdiff
1 11957_20230720   6.3005   2 2023-07-20 09:41:48 2023-07-20 13:39:40 14271.81
2 11957_20230720   6.3005   1 2023-07-20 09:35:42 2023-07-20 13:33:27 14265.51
3 11958_20230801  15.1875   2 2023-08-01 13:30:55 2023-08-01 17:28:50 14274.54
4 11958_20230801  15.1875   1 2023-08-01 13:24:59 2023-08-01 17:22:38 14259.35
5 11961_20230725  72.8050   1 2023-07-25 15:40:10 2023-07-25 19:37:53 14262.88
6 11961_20230725  72.8050   1 2023-07-25 15:38:58 2023-07-25 19:37:53 14335.69