Meeting 4: Summary and slides (final meeting): Difference between revisions
From Brain Mapping Unit
Jump to navigationJump to search
(Created page with "Meeting 4 took place on the19th of June 2012 ==Presentations & Discussion== ===Ameera: Latest Analyses=== Ameera showed results on three separate datasets: Carsten's data (nic...") |
|||
| Line 4: | Line 4: | ||
===Ameera: Latest Analyses=== | ===Ameera: Latest Analyses=== | ||
Ameera showed results on three separate datasets: Carsten's data (nicotine vs placebo), Karen and Simon's data (drug users) and Jonathan Power's data (children). | Ameera showed results on three separate datasets: Carsten's data (nicotine vs placebo), Karen and Simon's data (drug users) and Jonathan Power's data (children). The last two datasets can be found in the fMRI/Benchmark data section of this wiki. | ||
Ameera's slides can be found at: | Ameera's slides can be found at: | ||
[[http://bcni.psychol.cam.ac.uk/~pv226/Motion_Ameera_2.pdf]] | [[http://bcni.psychol.cam.ac.uk/~pv226/Motion_Ameera_2.pdf]] | ||
In a nutshell, she has shown that: | |||
*Carsten's data (as many other datasets with adults scanned at 3T) has much less motion than the Power dataset (children at 3T). Her diagnostics package can be used to show this for individual datasets. | |||
*Importantly, even in datasets with extensive motion the artifact described in the Power paper is *entirely* due to the pre-processing and can therefore be dealt with by careful pre-processing of the data (rater than scrubbing). | |||
*In practice, the last two points combined mean that many datasets that our group has worked on were probably immune from the artifact because of low levels of motion, careful preprocessing or both. There may be some exceptions to this, and it is recommended that people working of resting state fMRI data use the diagnostics package (available soon) for quality control. | |||
*The key aspect of preprocessing which seems to cause the artifact observed by Power et al is the ordering of the bandpass and regression steps. The effect produced by incorrect ordering is about half the size published by Power et al. Ameera is still investigating what may have compounded this effect in the substantially different pipeline used by Power et al. | |||
===Tiago: The issue of lag=== | ===Tiago: The issue of lag=== | ||
Revision as of 09:03, 20 June 2012
Meeting 4 took place on the19th of June 2012
Presentations & Discussion
Ameera: Latest Analyses
Ameera showed results on three separate datasets: Carsten's data (nicotine vs placebo), Karen and Simon's data (drug users) and Jonathan Power's data (children). The last two datasets can be found in the fMRI/Benchmark data section of this wiki.
Ameera's slides can be found at: [[1]]
In a nutshell, she has shown that:
- Carsten's data (as many other datasets with adults scanned at 3T) has much less motion than the Power dataset (children at 3T). Her diagnostics package can be used to show this for individual datasets.
- Importantly, even in datasets with extensive motion the artifact described in the Power paper is *entirely* due to the pre-processing and can therefore be dealt with by careful pre-processing of the data (rater than scrubbing).
- In practice, the last two points combined mean that many datasets that our group has worked on were probably immune from the artifact because of low levels of motion, careful preprocessing or both. There may be some exceptions to this, and it is recommended that people working of resting state fMRI data use the diagnostics package (available soon) for quality control.
- The key aspect of preprocessing which seems to cause the artifact observed by Power et al is the ordering of the bandpass and regression steps. The effect produced by incorrect ordering is about half the size published by Power et al. Ameera is still investigating what may have compounded this effect in the substantially different pipeline used by Power et al.
Tiago: The issue of lag
Tiago's slides will soon be available here.
Outcomes
We are please that the motion task force seems to have successfully dealt with the motion-related issues highlighted by recent papers such as Power et al 2012. The task-force is now dissolved, and the next steps will be as follows:
- Ameera will release a Beta version of her diagnostics toolbox (hopefully within the next week)
- The new BMU preprocessing pipeline will be distributed (over the next couple of months).
- We will have a joint meeting with the CBU, CamCAN and any other interested parties to disseminate these results in the first week of September. We will also set up a website and all relevant documentation for these packages by that time.
