View source for FBIRN:PhaseIIDiscussion

Phase II-- Do we need the repeatability for all 15 subjects?? The first 5 of each group may be sufficient; but to look at subject variability we may need all subjects. We're pretty sure that five will not be enough. 5 patients, 5 controls, 2 scans each = 20 scans as rapidly as possible. How much can be done in the next 8 weeks?

Clinical vs scans for the first 5 Sz and 5 controls--some sites are running out of clinician time; we need to be as good and as consistent as possible for review and publications. For sure we need the SANS/SAPS, done within a week of the scan. We need to try to be as complete as possible, but imaging is the most important dataset.

How many do we need? There is an algorithm: It needs the effect size and the ICC range you'll expect and can tell you the number of data points. But the site effects are still unknown.

'''At least two tracks'''<nowiki>: </nowiki>

Phase I publications

and Phase II data analysis and publications;

and the need for a soup-to-nuts imaging analysis pipeline that functions at all sites.

1) No one has published the actual *findings* of intersite variability per se.

2) No one has published on recommendations for scanner changes and comparisons.

3) Generalizability: Hal and Greg are re-doing their analyses. Greg needs summary statistics from specific ROIs.

Doug's Phase I masks: STG, primary motor, corpus callosum, primary visual areas (cortically defined), in subject's native space, as defined by the Freesurfer segmentation on the T1 data collected at MGH. We can use these masks to get summary z-scores etc.

'''what are the measures we want extracted??'''

For each contrast on those pre-defined ROI's:

1) Average Beta across the region (all values).

2) and the average of top 10% values (positive Beta's)

3) Ditto for the z scores

4) Ditto for the % signal change

'''What tasks?''' SM and BH; and MMN (constant tone blocks, mixed tone blocks, mixed > constants)--no point in doing SIRP at this moment.

Hal and Greg are working on SM and BH. Jess wants BH data, as does Gary.

''Lee needs precise slice-timing information from all sites!!''

<br />'''Imaging pipeline:'''

Need the scripts done; the event lumping method; way to store and upload the FLAC in the SRB and database; ability to sync the SRB and the local HID; etc. (from Doug's slides).

The common directory hierarchy is key, which the upload to the SRB will do.

Need organizing person at each site: IT people and data analysts. IT group is writing script to clarify the hierarchy; Need Syam's script to do Eprime Lumpring (Eprime2XML); Doug modify FIPS to work with correct dir. structure; each site to put data into correct hierarchy and run FIPS.

'''Phase II data analysis:''' Low-hanging fruit: Smoothing to a common level in order to combine data from multiple sites. We found a site x subject interaction in the Phase I data. Greg has the resources (Alex) to take a smoothing approach to the oddball to look for patient/control differences across sites. Send Greg the Auditory Oddball data from roughly a dozen datasets (Cindy, Jess, Judy, George, John).

All sites sending Aud Odd data to Greg: Get the BIAC/FBIRN QA tools from Syam and run them on the data, and send the movement/spikes info along with the data to Greg so he can decide whether or not to use the data. (This also demonstrates use of the FBIRN tools.)

He will need slice timing info and precise timing data from each site (in the E-prime) and will try to use the FBIRN FIPS pipeline.

Judy: Correlation maps with clinical scales; the combination across sites could be more powerful if the intersite difference removal works.