Presentation is loading. Please wait.

Presentation is loading. Please wait.

FBIRN Image Processing Stream (FIPS). Overview Goals First Level Analysis Steps FLAC Higher/Group Level Analysis Near-term Workflow.

Similar presentations


Presentation on theme: "FBIRN Image Processing Stream (FIPS). Overview Goals First Level Analysis Steps FLAC Higher/Group Level Analysis Near-term Workflow."— Presentation transcript:

1 fBIRN Image Processing Stream (FIPS)

2 Overview Goals First Level Analysis Steps FLAC Higher/Group Level Analysis Near-term Workflow

3 Long-Range Goals Queryable Highly flexible Multi-site Calibration Custom software plug-ins Prepackaged analysis software plug-ins Portalized

4 Short-Range Goals Limited flexibility FSL-only Publishable Start to solve data management problems

5 First Level (Run) Analysis Slice-timing Correction (?) Motion Correction Spatial Smoothing Intensity Normalization (Grand Mean) Fit General Linear Model (GLM) Regression Coefficients Residual Variance Contrasts of Regression Coefficients (Var) t, F, z maps of contrasts Clusters (GRF) Spaces: Native, Talairach (FSL-MNI152) ROI analysis/export

6 First-Level Analysis Results

7 FLAC – First-Level Analysis Configuration All pre-processing options (MC, slice timing, smoothing) Hemodynamic model (SPM HRF, delay, dispersion, derivative) Stimulus identity and timing (“Event Lumping”) Nuisance regressors Contrasts (would be better to do this later) Data set independent! Define ONCE!

8 Higher/Group Level Analysis Select First Level Analysis (FLAC) Select Contrast from First Level Select Subjects (Query) Group Model and Group Contrasts (GLM) Talairach Space (FSL-MNI152) Fit GLM Regression Coefficients Residual Variance Contrasts of Regression Coefficients t, z maps of contrasts Clusters (GRF)

9 Short-term Workflow “Local” Data Management Model Uses (Evolving) Hierarchy All data on SRB All data mirrored locally All analyses computed locally

10 First Level Analysis Workflow 1.Collect data for a session 2.Upload to session to SRB/HID 3.Synchronize LocalDB to SRB 4.Create stimulus timing files from E-Prime 5.Analyze (First Level) locally 6.Synchronize SRB to LocalDB (Upload) } Hierararchy Group Level Analysis Workflow -- TBD 0. Define and upload FLAC (once for all sessions) For each session:

11 Analyzing Data: fips-fla USAGE: fips-fla --flac FLACName : name of the FLAC as created with flac-checkin --subj SubjId : BIRN Subject ID (eg, 00900000103) --site SiteID : BIRN Site ID (eg, 006) --visit VisitNo : Visit number (eg, 1) --study StudyId : Study Id or number (eg, 1) --exp ExpId : Name of experiment (eg, sm) --db database : BIRN database: (eg, Human. Default from RC) --proj ProjectName : BIRN project (eg, fBIRNHumanPhantom. Default from RC) --projid ProjectID : BIRN project ID (eg, 000005. Default from RC) --overwrite : overwrite previously analyzed results --help : a short story about a Spanish boy named ‘Manual’ Processes all runs in experiment

12 Phase I Data Analyzed and on SRB Motion Correction Smoothing with FWHM=5mm Intensity Normalized BH, SM, and Rest – 15s ON, 15s OFF SIRP and MMN http://www.nmr.mgh.harvard.edu/~greve/fbirn/phase1

13

14 Database Analysis Parameters (FLAC) Session1 Data --Raw Pixel Data --Stimulus Id and Timing --Analyzed Data Session2 Data --Raw Pixel Data --Stimulus Id and Timing --Analyzed Data Site Data --Scanning protocol --Slice order --Skip

15 Database Analysis Parameters (FLAC) Session1 Data --Raw Pixel Data --Stimulus Id and Timing --Analyzed Data Session2 Data --Raw Pixel Data --Stimulus Id and Timing --Analyzed Data Site Data --Scanning protocol --Slice order --Skip Create Analysis (FLAC)

16 Configuring the First-Level Analysis using FSL-FEAT

17

18

19

20 FLAC – First-Level Analysis Configuration Save to create the configuration file Import/upload Data set independent! Define ONCE!

21 Stimulus Identification and Timing 17342568 HAPPYSAD Happy.stf: 10.0 2 1 30.0 2 1 28.7 6 1 43.1 2 1 Sad.stf: 3.0 2 1 13.0 2 1 15.0 2 1 37.1 2 1 Time t=0t=3.0t=10.0 t=13.0 t=15.0t=30.0t=28.7t=37.1t=43.1 Columns: 1.Onset Time 2.Duration 3.Weight Example: Happy and Sad Faces as defined by the experimenter.

22 Stimulus Identification and Timing 17342568 HAPPYSAD Happy.stf: 10.0 2 1 30.0 2 1 28.7 6 1 43.1 2 1 Sad.stf: 3.0 2 1 13.0 2 1 15.0 2 1 37.1 2 1 Subject 17342568 Experimenter HAPPYSAD HappySubj.stf: 10.0 2 1 15.0 2 1 30.0 2 1 43.1 2 1 SadSubj.stf: 3.0 2 1 13.0 2 1 28.7 6 1 37.1 2 1 ExperimenterSubject

23 17342568 17342568 17342568 HHHS SHSS Subject HS H S Experimenter SH.stf: 15.0 2 1 HH.stf: 10.0 2 1 28.7 6 1 43.1 2 1 HS.stf: 28.7 6 1 SS.stf: 3.0 2 1 13.0 2 1 37.1 2 1

24 Analyze Specify FLAC and Session Session indicates: where the data are site (slice order, skip, …) FLAC indicates: relative names of raw data, timing files, etc FLAC customized (full paths resolved, etc) Design matrix created GLM Estimation Inference (contrasts, z, t) Resample to Talairach space Local: Upload/Synch with SRB Database Analysis Parameters (FLAC) Session Data --Raw Pixel Data --Stimulus Id and Timing --Analyzed Data Session Data --Raw Pixel Data --Stimulus Id and Timing --Analyzed Data Site Data --Scanning protocol --Slice order --Skip

25 Group Analysis Group Level Analysis Configuration (GLAC?) Specify: FLAC as input FLAC contrast Group model (eg, Age, IQ, diagnosis, etc) Group contrasts Can be done within FSL

26 Where we are now: Local Analysis (ie, not on the SRB/Portal) Scripts to define FLAC and analyze data Data must be in proper hierarch – but no way to achieve this easily (need upload script)

27 Immediate Goals: Finalize session hierarchy Add XML to FSL output directory Resolve stimulus ID and timing specification Resolve site-dependent parameters Update FIPS scripts Local Analysis with SRB synch (?)

28 Intermediate Goals: Portalize and/or incorporate into upload script Group analysis Long-Rang Goals: Generalize FLAC (ie, not dependent upon FSL) Incorporate other analysis packages Incorporate Calibration Recommendations

29

30 Data Hierarchy: Image Data VARIABLE NAMES ------------------------------------ LOCAL_SRB_HOME_BIRN/ DataBase/ ProjectName_ProjectId SubjectID/ Visit__SiteId_VisitNo/ Study__StudyId/ Experiment__Series/ Native/ Original__0001/ Analyze/ EXAMPLE -------------------------------------- /home/BIRN/ Human/ fBIRNHumanPhantom__000005/ 00900000103/ Visit__009_0001/ Study__0001/ sm__0003/ Native/ Original__0001/ Analyze/ ---- RC CL -- RC = ~/.fipsrc CL = Command-line In Analyze/: f0001.img, f0002.img, …

31 FIPS RC File (~/.fipsrc) setenv LOCAL_SRB_HOME_BIRN /space/birn/26/users/greve/srb-mirror setenv FIPS_FLAC_DIR ~/flac setenv FIPS_DEFAULT_DATABASE Human setenv FIPS_DEFAULT_PROJECT_NAME fBIRNHumanPhantom setenv FIPS_DEFAULT_PROJECT_ID 000005

32 Analyzing Data: fips-fla USAGE: fips-fla --flac FLACName : name of the FLAC as created with flac-checkin --subj SubjId : BIRN Subject ID (eg, 00900000103) --site SiteID : BIRN Site ID (eg, 006) --visit VisitNo : Visit number (eg, 1) --study StudyId : Study Id or number (eg, 1) --exp ExpId : Name of experiment (eg, sm) --db database : BIRN database: (eg, Human. Default from RC) --proj ProjectName : BIRN project (eg, fBIRNHumanPhantom. Default from RC) --projid ProjectID : BIRN project ID (eg, 000005. Default from RC) --overwrite : overwrite previously analyzed results --help : a short story about a Spanish boy named ‘Manual’ Processes all runs in experiment

33 Data Hierarchy: FSL Output VARIABLE NAMES ------------------------------------ LOCAL_SRB_HOME_BIRN/ DataBase/ ProjectName_ProjectId SubjectID/ Visit__SiteId_VisitNo/ Study__StudyId/ Experiment__Series/ Analysis/ Original__0001/ FLACNAME.feat/ EXAMPLE -------------------------------------- /home/BIRN/ Human/ fBIRNHumanPhantom__000005/ 00900000103/ Visit__009_0001/ Study__0001/ sm__0003/ Analysis/ Original__0001/ sensorimotor.feat/ ---- RC CL -- RC = ~/.fipsrc CL = Command-line In TaskData/ sensorimotor.stf

34 Creating a FLAC 1.Create temporary dir 2.Copy one run of raw data (in analyze4d) - template 3.Copy timing files (if using) – name important 4.Run FSL Feat GUI (specify processing options) – do not actually analyze data. 5.flac-checkin (FIPS_FLAC_DIR) 6.Delete temporary dir 7.Do this once for each parameter set 8.Analyze all data with this FLAC flac-checkin --help

35 To Analyze Data 1.Copy data into local SRB mirror 2.Copy timing files into TaskData/ for each run. Use the same names as when defining the FLAC. Names stay the same across runs, subjects, etc. 3.Analyze with fips-fla, which will customize the FLAC for each data set 4.View with html browser or AFNI

36 Data Hierarchy: Timing Files VARIABLE NAMES ------------------------------------ LOCAL_SRB_HOME_BIRN/ DataBase/ ProjectName_ProjectId SubjectID/ Visit__SiteId_VisitNo/ Study__StudyId/ Experiment__Series/ Native/ Original__0001/ TaskData/ EXAMPLE -------------------------------------- /home/BIRN/ Human/ fBIRNHumanPhantom__000005/ 00900000103/ Visit__009_0001/ Study__0001/ sm__0003/ Native/ Original__0001/ TaskData/ ---- RC CL -- RC = ~/.fipsrc CL = Command-line In TaskData/ sensorimotor.stf

37 Timing Files (FSL: 3-column custom) 15 15 1 45 15 1 75 15 1 105 15 1 135 15 1 165 15 1 195 15 1 22515 1 Onset Time (sec), Duration (sec), Weight One timing file for each event type.

38 Outstanding Issues How to upload timing files to SRB How to upload FLAC to SRB Adding XML files to FSL output Synching output with SRB Slice-timing correction – site dependent Delete/Skip first few volumes – site dependent

39 Future FLAC creation without a template data set All processing through portal – no need to have local mirror Interface with group analysis Generalize FLAC to other packages

40

41

42 First Level Analysis Workflow 1.Collect data for a session 2.Upload to session to SRB/HID 3.Synchronize LocalDB to SRB 4.Create stimulus timing files from E-Prime 5.Analyze (First Level) locally 6.Synchronize SRB to LocalDB (Upload) “Local” Data Management Model All data on SRB All data mirrored locally All analyses computed locally } Hierararchy


Download ppt "FBIRN Image Processing Stream (FIPS). Overview Goals First Level Analysis Steps FLAC Higher/Group Level Analysis Near-term Workflow."

Similar presentations


Ads by Google