Using 3dMSS with a continuous input variable

rclements · October 12, 2023, 3:22pm

Hello,

I would like to use 3dMSS to model the response shape to changes in end-tidal CO2. Is it possible to use a continuous input like an end-tidal CO2 regressor with 3dMSS? If so, how would I do this?

Thank you,
Becca

Gang · October 12, 2023, 5:25pm

Becca,

Could you provide further details about your data structure? Is the data organized at the individual level or population level? Is end-tidal CO2 measured on a per-trial basis? Additionally, considering end-tidal CO2 as the predictor, could you clarify what the response variable is?

Gang

rclements · October 16, 2023, 4:35pm

Hi Gang,

Thanks for your response. For each scan, we collect end-tidal CO2 data at a sampling rate of 1000 Hz. If necessary, this data can be downsampled to the number of TRs (so that for each TR there is a corresponding end-tidal CO2 value). During each scan, the participants did the same breath-hold task which causes large changes in end-tidal CO2.

My original goal was just to estimate individual breath-hold BOLD response shapes for each scan based on that subject/scan's unique end-tidal CO2 trace. I found these instructions (https://docs.google.com/document/d/1ksnsT8y0a-nDasdHlfZi4ytVKrfoJ-TZff-rZXG6nuM/edit) which describe exactly what I wanted to do, but my estimated response functions do not look reasonable (see example below).

Although estimating individual/scan level responses would be ideal, it seems like doing some sort of group level analysis may be necessary to get more reasonable looking HRFs, which is why I was trying to use 3dMSS. I have data for a group of 7 participants scanned 10 times each (during each scan, there was the same breath-hold task and there is a different end-tidal CO2 trace for each of these 70 scans).

These scans all took place within a short time period and the participants are healthy, so I am not interested in any other variables other than just what the BOLD response shape to the breath-hold task is. Would incorporating the scan specific end-tidal CO2 regressor into 3dMSS be possible to estimate population level HRFs?

Thank you!
Becca

Gang · October 17, 2023, 1:37am

Becca,

Regarding the data collection, for each scan, we capture end-tidal CO2 data at a rate of 1000 Hz. If needed, we can downsample this data to match the number of TRs, ensuring there's a corresponding end-tidal CO2 value for each TR. During every scan, participants perform the same breath-hold task, which leads to significant fluctuations in end-tidal CO2.

Some clarifications:

When you mention a "scan," are you referring to a single breath-hold episode, or does it encompass multiple episodes? Additionally, could you specify the duration of each breath-hold in terms of TRs?
Regarding the estimated BOLD response that don't appear reasonable, are these estimations based on a single or multiple breath-hold episodes for an individual?

Would incorporating the scan specific end-tidal CO2 regressor into 3dMSS be possible to estimate population level HRFs?

Considering your interest in incorporating scan-specific end-tidal CO2 regressors into 3dMSS for estimating population-level HRFs, it's indeed possible. 3dMSS can be employed to estimate the BOLD response at either the individual or population level, depending on the research objectives. Once I have a clearer understanding of the aforementioned questions, I can provide more details.

Gang

rclements · October 17, 2023, 2:14pm

Hi Gang,

Each scan consists of 6 breath-hold trials - a trial consists of 24 seconds paced breathing, a 15 second breath-hold, a 2 second exhale, and an 8 second recovery period. There is also one more round of paced breathing at the end of the 6 trials. My TR is 1.5 seconds so the breath-hold itself is 10 TRs, but the whole scan is 212 TRs.

My estimated BOLD responses that don't appear reasonable are based on the data for the entire scan (so 6 breath-hold trials). I was using the end-tidal CO2 trace for the entire scan as well as the imaging data for the entire scan to calculate the BOLD response using the instructions I linked above.

Thank you for your help!

Becca

Gang · October 18, 2023, 2:21am

Becca,

A further clarification: are you aiming (1) to assess the average BOLD response magnitude (i.e., a single number) across the six trials, (2) to capture the average BOLD response profile (over 10 TRs of breath-hold) across those six trials, or (3) to estimate the BOLD response from the beginning to the end of each scan?

Could you provide the specific parameter values (e.g., dur_tent, npar_tent) you used for the following line in the Google Doc?

-stim_times_AM1 1 CCttkk.1D "TENT(0,${dur_tent},${npar_tent})" \

Additionally, could you share the content of the file "CCttkk.1D"?

BTW, do all participants share the same breath-hold timing?

Gang

rclements · October 18, 2023, 4:11pm

Hi Gang,

I am interested in assessing the average BOLD response profile over the breath-hold across the 6 trials. We expect the effects of the breath-hold to last longer than 10 TRs though and also want to account for the fact that different regions of the brain have different onset delays. To account for this, I want the duration of the tent to be 20-30 seconds. So for each scan, I just want to get one response shape that is 20-30 seconds long (this response shape might have an onset delay of 0-10 seconds depending on the brain region).

I can email my "CCttkk.1D" file and also my script. I was using TENT(0,20,11) to get the estimated response functions that weren't looking reasonable. I realized that if I switch to TENTzero(0,20,5) I can get reasonable looking response functions, but since there are only 3 parameters being estimated per voxel here, so I'm not able to capture the onset delay. I would prefer to use more tent functions to better capture the full response shape and onset delay.

All participants do share the same breath-hold timing.

Thank you!