Difference between revisions of "Suggested Pipelines"

From MEG Core
Jump to navigation Jump to search
 
(45 intermediate revisions by the same user not shown)
Line 13: Line 13:
   
 
{{#mermaid:graph LR
 
{{#mermaid:graph LR
  +
subgraph MRI Preprocessing
 
 
MRI["Structural MRI"] --> fids["Place Fiducials"];
 
MRI["Structural MRI"] --> fids["Place Fiducials"];
 
MRI["Structural MRI"] --> shape["Register point clouds"];
 
MRI["Structural MRI"] --> shape["Register point clouds"];
  +
  +
subgraph MRI Preprocessing
  +
 
fids --> hull["Process with orthohull"];
 
fids --> hull["Process with orthohull"];
 
shape --> hull;
 
shape --> hull;
Line 35: Line 38:
 
== Basic Resting State MEG processing ==
 
== Basic Resting State MEG processing ==
   
When examining resting state data, the end goals is usually to examine either static measures of power, or connectivity. For connectivity, the output of SAM is a continuous time series, usually the Hilbert envelope of a band limited signal. Following calculation of this time series, other routines (such as ICA, seed based correlation, etc.) can be used to derive connectivity between regions.
+
Basic preprocessing of resting state MEG data includes filtering, and possibly artifact removal. Removing artifacts could consist of eliminating bad segments, or a more comprehensive process like ICA. When examining resting state data, the end goals is usually to examine either static measures of power, or connectivity. For connectivity, the output of SAM is a continuous time series, usually the Hilbert envelope of a band limited signal. Following calculation of this time series, other routines (such as ICA, seed based correlation, etc.) can be used to derive connectivity between regions.
   
 
{{#mermaid:graph LR
 
{{#mermaid:graph LR
  +
MEG["MEG Data"] --> Filtering;
 
 
subgraph PreProcessing
 
subgraph PreProcessing
MEG["MEG Data"] --> Filtering;
 
 
Filtering --> Art["Artifact Removal"];
 
Filtering --> Art["Artifact Removal"];
  +
end
  +
  +
subgraph SAM PreProcessing
  +
Art --> sam_cov
  +
sam_cov --> sam_wts
  +
style sam_cov fill:#fcf
  +
style sam_wts fill:#fcf
  +
click sam_cov "https://megcore.nih.gov/index.php/Sam_cov" "sam_cov documentation"
  +
click sam_wts "https://megcore.nih.gov/index.php/Sam_wts" "sam_wts documentation"
 
end
 
end
   
 
subgraph Power
 
subgraph Power
Art --> sam["sam_3d/sam_3dc"];
+
sam_wts --> sam["sam_3d/sam_3dc"];
   
 
style sam fill:#fcf
 
style sam fill:#fcf
Line 51: Line 62:
 
end
 
end
 
subgraph Connectivity
 
subgraph Connectivity
Art --> sam_power;
+
sam_wts --> sam_power;
 
style sam_power fill:#fcf
 
style sam_power fill:#fcf
 
click sam_power "https://megcore.nih.gov/index.php/Sam_power" "sam_power documentation"
 
click sam_power "https://megcore.nih.gov/index.php/Sam_power" "sam_power documentation"
 
end
 
end
 
}}
 
}}
  +
  +
==Basic Task Based MEG Pipeline==
  +
  +
In a task based analysis, you start with raw MEG data, as wells as data from the ADC channels - triggers, stimuli, and responses. Both of these must be pre-processed. Once your MEG data is marked appropriately, you can begin a SAM analysis. If you are interested in time-locked (evoked or event-related) signals, you can use either [[sam_4d_and_sam_4dc|sam_4d or sam_4dc]] or [[sam_ers_and_sam_ersc|sam_ers or sam_ersc]], depending on exactly what you want as output. Alternatively, if you do not expect your signals to be time-locked, you can examine changes in induced power using [[sam_3d_and_sam_3dc|sam_3d or sam_3dc]].
   
 
{{#mermaid:graph LR
 
{{#mermaid:graph LR
  +
MEG["MEG Data"] --> Filtering;
   
  +
subgraph MEG PreProcessing
marks --> Covariance;
 
  +
Filtering --> Art["Artifact Removal"];
  +
end
   
  +
ADC["ADC/Trigger Channels"] --> thresholdDetect;
subgraph MEG Preprocessing
 
  +
raw[Raw MEG data] --> filter[Basic Filtering];
 
  +
subgraph ADC PreProcessing
adc[Raw ADC/PPT<br>data] --> ThresholdDetect;
 
  +
thresholdDetect --> add_markers;
ThresholdDetect --> marks[Create Markers];
 
filter --> meg[MEG Data];
 
 
end
 
end
   
subgraph MEG Data Statistics
+
subgraph SAM PreProcessing
  +
add_markers--> sam_cov;
meg --> Filter[Band-pass<br>filter];
 
Filter --> Covariance;
+
Art --> sam_cov;
  +
sam_cov --> sam_wts;
  +
style sam_cov fill:#fcf
  +
style sam_wts fill:#fcf
  +
click sam_cov "https://megcore.nih.gov/index.php/Sam_cov" "sam_cov documentation"
  +
click sam_wts "https://megcore.nih.gov/index.php/Sam_wts" "sam_wts documentation"
 
end
 
end
   
  +
subgraph Induced Power
  +
sam_wts --> sam_3d["sam_3d/sam_3dc"];
  +
style sam_3d fill:#fcf
  +
click sam_3d "https://megcore.nih.gov/index.php/sam_3d_and_sam_3dc" "sam_3d/sam_3dc documentation"
  +
end
  +
  +
subgraph Evoked/Event Related
  +
sam_wts --> sam_ers["sam_ers/sam_ersc"];
  +
sam_wts --> sam_4d["sam_4d/sam_4dc"];
  +
style sam_ers fill:#fcf
  +
click sam_ers "https://megcore.nih.gov/index.php/sam_ers_and_sam_ersc" "sam_ers/sam_ersc documentation"
  +
style sam_4d fill:#fcf
  +
click sam_4d "https://megcore.nih.gov/index.php/sam_4d_and_sam_4dc" "sam_4d/sam_4dc documentation"
  +
end
 
}}
 
}}
  +
  +
== Localizing Epileptiform Activity ==
  +
  +
This describes localizing epileptiform activity from continuous datasets using excess kurtosis. First, a SAM analysis is performed with a relatively narrow bandwidth, 20-70Hz. The program sam_epi is used to produce images of kurtosis, and NIFTIpeak is used to find extrema in the images that may indicate voxels which contain spikes. Once targets are identified, the SAM analysis is essentially run again, this time with a wide bandwidth, on only those specific targets. DataEditor can read the weights for those targets and compute "virtual sensors" at those voxel locations. The epileptologist can then examine the time series and mark epileptic spikes.
   
 
{{#mermaid:graph LR
 
{{#mermaid:graph LR
  +
subgraph Synthetic Aperture Magnetometry
 
  +
MEG
Covariance --> Beamformer;
 
  +
head[Head Model] --> Beamformer;
 
Beamformer --> image["3D Images"];
+
MEG["MEG Data"] --> sam_cov;
  +
  +
subgraph Phase I: Filter 20-70Hz
  +
sam_cov --> sam_wts;
  +
sam_wts --> sam_epi;
  +
style sam_cov fill:#fcf
  +
click sam_cov "https://megcore.nih.gov/index.php/Sam_cov" "sam_cov documentation"
  +
style sam_wts fill:#fcf
  +
click sam_wts "https://megcore.nih.gov/index.php/Sam_wts" "sam_wts documentation"
 
end
 
end
  +
  +
sam_epi --> NIFTIPeak["NIFTIPeak <br>Target <br>Identification"]
  +
NIFTIPeak --> sam_cov2["sam_cov"]
  +
style sam_epi fill:#fcf
  +
click sam_epi "https://megcore.nih.gov/index.php/sam_epi" "sam_epi documentation"
  +
style NIFTIPeak fill:#fcf
  +
click NIFTIPeak "https://megcore.nih.gov/index.php/NIFTIPeak" "NIFTIPeak.py documentation"
  +
  +
subgraph Phase II: Filter 1-150Hz
  +
sam_cov2 --> sam_wts2["sam_wts"]
  +
style sam_cov2 fill:#fcf
  +
click sam_cov2 "https://megcore.nih.gov/index.php/Sam_cov" "sam_cov documentation"
  +
style sam_wts2 fill:#fcf
  +
click sam_wts2 "https://megcore.nih.gov/index.php/Sam_wts" "sam_wts documentation"
  +
end
  +
sam_wts2 --> DataEditor["DataEditor <br> Spike Identification"]
  +
 
}}
 
}}
  +
  +
== Advanced Coregistration SAM Pipeline ==
  +
  +
There is a limit to the accuracy with which MEG and MRI data can be coregistered using fiducial markers alone. One significant confounding issue is that the brain will be in a slightly different position inside the head in a seated position vs. a supine position. This pipeline describes using sam_coreg to refine the approximate fiducial placements. Once sam_coreg has been performed, the user can either do a traditional analysis using the SAM tools discussed previously, or make use of the programs that leverage the cortical data (patch_wts and roi_wts).
   
 
{{#mermaid:graph LR
 
{{#mermaid:graph LR
subgraph SAM Workflow
 
sam_cov --> sam_wts;
 
sam_wts --> sam_3d;
 
sam_3d --> AFNI;
 
AFNI --> sam_wts;
 
AFNI --> sam_cov;
 
   
  +
MRI["Structural MRI"] --> fids["Place Approximate <br> Fiducials"];
  +
  +
subgraph MRIproc
  +
fids --> FreeSurfer
  +
FreeSurfer --> FSNormals.py
  +
style FSNormals.py fill:#fcf
  +
click FSNormals.py "https://megcore.nih.gov/index.php/FSNormals.py" "FSNormals.py documentation"
  +
end
  +
  +
FSNormals.py --> sam_coreg
  +
MEG-->sam_cov
 
style sam_cov fill:#fcf
 
style sam_cov fill:#fcf
click sam_cov "https://megcore.nih.gov/index.php/Sam_cov" "sam documentation"
+
click sam_cov "https://megcore.nih.gov/index.php/Sam_cov" "sam_cov documentation"
  +
  +
subgraph MEGproc
  +
sam_cov-->sam_coreg
  +
sam_coreg--> patch_wts
  +
sam_coreg--> roi_wts
  +
sam_coreg--> sam_wts
  +
style sam_coreg fill:#fcf
  +
click sam_coreg "https://megcore.nih.gov/index.php/sam_coreg" "sam_coreg documentation"
  +
style patch_wts fill:#fcf
  +
click patch_wts "https://megcore.nih.gov/index.php/patch_wts" "patch_wts documentation"
  +
style roi_wts fill:#fcf
  +
click roi_wts "https://megcore.nih.gov/index.php/roi_wts" "roi_wts documentation"
 
style sam_wts fill:#fcf
 
style sam_wts fill:#fcf
click sam_wts "https://megcore.nih.gov/index.php/Sam_wts" "sam documentation"
+
click sam_wts "https://megcore.nih.gov/index.php/sam_wts" "sam_wts documentation"
  +
end
  +
  +
subgraph SAM
  +
sam_wts-->sam_3d["sam_3d or sam_3dc"]
  +
sam_wts-->sam_4d["sam_4d or sam_4dc"]
  +
sam_wts-->sam_ers["sam_ers or sam_ersc"]
  +
sam_wts-->sam_power["sam_power"]
 
style sam_3d fill:#fcf
 
style sam_3d fill:#fcf
click sam_3d "https://megcore.nih.gov/index.php/Sam_3d" "sam documentation"
+
click sam_3d "https://megcore.nih.gov/index.php/sam_3d_and_sam_3dc" "sam_3d documentation"
style AFNI fill:#fcf
+
style sam_4d fill:#fcf
click AFNI "https://afni.nimh.nih.gov/" "The AFNI website"
+
click sam_4d "https://megcore.nih.gov/index.php/sam_4d_and_sam_4dc" "sam_4d documentation"
  +
style sam_ers fill:#fcf
  +
click sam_ers "https://megcore.nih.gov/index.php/roi_wts" "sam_ers documentation"
  +
style sam_power fill:#fcf
  +
click sam_power "https://megcore.nih.gov/index.php/sam_power" "sam_power documentation"
 
end
 
end
 
}}
 
}}
 
# Create covariance matrices using sam_cov.
 
# Compute beamformer weights with sam_wts.
 
# sam_3d uses the weights to compute volumetric images of activity estimates.
 
# View them with AFNI.
 
# It didn't work, go back and try again.
 
# Nope, still didn't work, try this instead.
 

Latest revision as of 15:29, 20 March 2019

Master Pipeline

Basic MRI Pre-Processing Workflow

For any experiment where you wish to localize data to the brain, the first step is MRI pre-processing. First, MEG data must be co-registered to the space of the MRI, either by manually placing fiducial points on the MRI, or through a semi-automated method where a digital head shape is aligned with a head surface. (Other algorithmic techniques are possible, these will be discussed later). For the purpose of source space reconstruction, the head can be modeled either as a collection of spheres, one per channel, (MultiSphere) or in a realistic fashion using the Nolte model.

Basic Resting State MEG processing

Basic preprocessing of resting state MEG data includes filtering, and possibly artifact removal. Removing artifacts could consist of eliminating bad segments, or a more comprehensive process like ICA. When examining resting state data, the end goals is usually to examine either static measures of power, or connectivity. For connectivity, the output of SAM is a continuous time series, usually the Hilbert envelope of a band limited signal. Following calculation of this time series, other routines (such as ICA, seed based correlation, etc.) can be used to derive connectivity between regions.

Basic Task Based MEG Pipeline

In a task based analysis, you start with raw MEG data, as wells as data from the ADC channels - triggers, stimuli, and responses. Both of these must be pre-processed. Once your MEG data is marked appropriately, you can begin a SAM analysis. If you are interested in time-locked (evoked or event-related) signals, you can use either sam_4d or sam_4dc or sam_ers or sam_ersc, depending on exactly what you want as output. Alternatively, if you do not expect your signals to be time-locked, you can examine changes in induced power using sam_3d or sam_3dc.

Localizing Epileptiform Activity

This describes localizing epileptiform activity from continuous datasets using excess kurtosis. First, a SAM analysis is performed with a relatively narrow bandwidth, 20-70Hz. The program sam_epi is used to produce images of kurtosis, and NIFTIpeak is used to find extrema in the images that may indicate voxels which contain spikes. Once targets are identified, the SAM analysis is essentially run again, this time with a wide bandwidth, on only those specific targets. DataEditor can read the weights for those targets and compute "virtual sensors" at those voxel locations. The epileptologist can then examine the time series and mark epileptic spikes.

Advanced Coregistration SAM Pipeline

There is a limit to the accuracy with which MEG and MRI data can be coregistered using fiducial markers alone. One significant confounding issue is that the brain will be in a slightly different position inside the head in a seated position vs. a supine position. This pipeline describes using sam_coreg to refine the approximate fiducial placements. Once sam_coreg has been performed, the user can either do a traditional analysis using the SAM tools discussed previously, or make use of the programs that leverage the cortical data (patch_wts and roi_wts).