• If you are citizen of an European Union member nation, you may not use this service unless you are at least 16 years old.

  • You already know Dokkio is an AI-powered assistant to organize & manage your digital files & messages. Very soon, Dokkio will support Outlook as well as One Drive. Check it out today!


Cluster Detection

Page history last edited by PBworks 17 years ago

(1) Start MEDx, and open a New Folder.


(2) Load in one of the Z maps that you created in either Analysis 1: Simple t-test or Analysis 2: Multiple Regression. For the purpose of this example, let's assume that the Z map was named Z.simple.RealWord-FalseFont.hdr.


(3) Load the Mask image that you created in step 12 of Spatial Normalization (the one named SnMask.hdr, not Mask.hdr).


(4) Select Toolbox --> Functional --> Final Significance...


(5) In the Final Assessment of Significance dialog box, click on the Cluster Detection tab.


(6) Set both Statistical Map and Estimate Source to the Z map you loaded in step 2, and Mask Volume to the mask you loaded in step 3. Use the Select... buttons if you wish.


(7) Under Method for Estimating Spatial Smoothness, leave the selection set to Optimized for Low Smoothness.


(8) Under Cluster Analysis Parameters, leave Z-Value Threshold set to 2.33, and leave Probability Threshold set to 0.05.


(9) Set output file to a file in your tutorial directory that encodes the parameters of this Cluster Detection method, e.g.




Note that this naming convention indicates that this is a Cluster Detection text file (the extension ends in .txt), generated from the Z map Z.simple.RealWord-FalseFont.hdr, using a Z threshold of 2.33 and p-value threshold of 0.05.


(10) Click on the OK button. This causes Cluster Detection to be performed on the Z map. (A Confirm box stating that the Z map does not appear to be a Z-Map volume, and asking whether you want to continue. Click on the Yes button. This is one of those design decisions in MEDx that I think was a mistake. At one point I thought I had removed this Confirm box from the Cluster Detection module while at Sensor Systems, but it looks like somebody put it back in -- probably accidentally.)


A new group containing two buttons labeled Cluster Mask and Probability Mask will be generated.


(11) Use gedit to inspect the contents of the new text file, e.g.


gedit Cluster.Z.Simple.RealWord-FalseFont._2.33._0.05.txt


At the bottom, statistically significant clusters (i.e., clusters whose p-values are below the p-value threshold you chose) are listed. If you used the Z map Z.simple.RealWord-FalseFont.hdr, you should see four clusters listed, of sizes 15301, 4135, 4048, and 1130 voxels, with p-values ranging from 3.33067e-16 to 0.0370248.


(12) Click on the button labeled Cluster Mask. This image is in floating point format (not TIFF!), but the Overlays step requires a cluster mask in 16-bit integer format. Select Toolbox --> Conversion --> Data Encoding... and change Output Pixel Encoding to Unsigned Integer and Output Depth to 16**. Then save the output image to the tutorial directory in AVW format, giving it a descriptive name, e.g.




(Note that I have used almost exactly the same name as the output text file, except that I changed that extension from .txt to .hdr.)


(13) Exit MEDx.


The next step is: Overlays.


For more information on Cluster Detection, see Section 31.13.3 of Chapter 31 of the MEDx User's Guide.


Return to tutorial main page

Comments (0)

You don't have permission to comment on this page.